diff --git "a/coreml_model.mlmodelc/model.mil" "b/coreml_model.mlmodelc/model.mil" new file mode 100644--- /dev/null +++ "b/coreml_model.mlmodelc/model.mil" @@ -0,0 +1,4284 @@ +program(1.0) +[buildInfo = dict, tensor>({{"coremlc-component-MIL", "3304.5.2"}, {"coremlc-version", "3304.6.2"}})] +{ + func main(tensor conditioning_scale, tensor controlnet_cond, tensor encoder_hidden_states, tensor sample, tensor text_embeds, tensor time_ids, tensor timestep) { + tensor var_45 = const()[name = tensor("op_45"), val = tensor(-1)]; + tensor var_61_axes_0 = const()[name = tensor("op_61_axes_0"), val = tensor([1])]; + tensor var_61_cast_fp16 = expand_dims(axes = var_61_axes_0, x = timestep)[name = tensor("op_61_cast_fp16")]; + tensor var_63_to_fp16 = const()[name = tensor("op_63_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(64)))]; + tensor emb_3_cast_fp16 = mul(x = var_61_cast_fp16, y = var_63_to_fp16)[name = tensor("emb_3_cast_fp16")]; + tensor var_68_cast_fp16 = sin(x = emb_3_cast_fp16)[name = tensor("op_68_cast_fp16")]; + tensor var_69_cast_fp16 = cos(x = emb_3_cast_fp16)[name = tensor("op_69_cast_fp16")]; + tensor emb_7_interleave_0 = const()[name = tensor("emb_7_interleave_0"), val = tensor(false)]; + tensor emb_7_cast_fp16 = concat(axis = var_45, interleave = emb_7_interleave_0, values = (var_68_cast_fp16, var_69_cast_fp16))[name = tensor("emb_7_cast_fp16")]; + tensor var_73_begin_0 = const()[name = tensor("op_73_begin_0"), val = tensor([0, 160])]; + tensor var_73_end_0 = const()[name = tensor("op_73_end_0"), val = tensor([1, 320])]; + tensor var_73_end_mask_0 = const()[name = tensor("op_73_end_mask_0"), val = tensor([true, true])]; + tensor var_73_cast_fp16 = slice_by_index(begin = var_73_begin_0, end = var_73_end_0, end_mask = var_73_end_mask_0, x = emb_7_cast_fp16)[name = tensor("op_73_cast_fp16")]; + tensor var_75_begin_0 = const()[name = tensor("op_75_begin_0"), val = tensor([0, 0])]; + tensor var_75_end_0 = const()[name = tensor("op_75_end_0"), val = tensor([1, 160])]; + tensor var_75_end_mask_0 = const()[name = tensor("op_75_end_mask_0"), val = tensor([true, false])]; + tensor var_75_cast_fp16 = slice_by_index(begin = var_75_begin_0, end = var_75_end_0, end_mask = var_75_end_mask_0, x = emb_7_cast_fp16)[name = tensor("op_75_cast_fp16")]; + tensor t_emb_interleave_0 = const()[name = tensor("t_emb_interleave_0"), val = tensor(false)]; + tensor t_emb_cast_fp16 = concat(axis = var_45, interleave = t_emb_interleave_0, values = (var_73_cast_fp16, var_75_cast_fp16))[name = tensor("t_emb_cast_fp16")]; + tensor time_embedding_linear_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(307712))), name = tensor("time_embedding_linear_1_weight_to_fp16_palettized"), shape = tensor([1280, 320])]; + tensor time_embedding_linear_1_bias_to_fp16 = const()[name = tensor("time_embedding_linear_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(307904)))]; + tensor linear_0_cast_fp16 = linear(bias = time_embedding_linear_1_bias_to_fp16, weight = time_embedding_linear_1_weight_to_fp16_palettized, x = t_emb_cast_fp16)[name = tensor("linear_0_cast_fp16")]; + tensor input_5_cast_fp16 = silu(x = linear_0_cast_fp16)[name = tensor("input_5_cast_fp16")]; + tensor time_embedding_linear_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(310528))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1539392))), name = tensor("time_embedding_linear_2_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor time_embedding_linear_2_bias_to_fp16 = const()[name = tensor("time_embedding_linear_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1539584)))]; + tensor linear_1_cast_fp16 = linear(bias = time_embedding_linear_2_bias_to_fp16, weight = time_embedding_linear_2_weight_to_fp16_palettized, x = input_5_cast_fp16)[name = tensor("linear_1_cast_fp16")]; + tensor concat_0 = const()[name = tensor("concat_0"), val = tensor([6])]; + tensor timesteps_cast_fp16 = reshape(shape = concat_0, x = time_ids)[name = tensor("timesteps_cast_fp16")]; + tensor var_95 = const()[name = tensor("op_95"), val = tensor(-1)]; + tensor var_111_axes_0 = const()[name = tensor("op_111_axes_0"), val = tensor([1])]; + tensor var_111_cast_fp16 = expand_dims(axes = var_111_axes_0, x = timesteps_cast_fp16)[name = tensor("op_111_cast_fp16")]; + tensor var_113_to_fp16 = const()[name = tensor("op_113_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1542208)))]; + tensor emb_11_cast_fp16 = mul(x = var_111_cast_fp16, y = var_113_to_fp16)[name = tensor("emb_11_cast_fp16")]; + tensor var_118_cast_fp16 = sin(x = emb_11_cast_fp16)[name = tensor("op_118_cast_fp16")]; + tensor var_119_cast_fp16 = cos(x = emb_11_cast_fp16)[name = tensor("op_119_cast_fp16")]; + tensor emb_15_interleave_0 = const()[name = tensor("emb_15_interleave_0"), val = tensor(false)]; + tensor emb_15_cast_fp16 = concat(axis = var_95, interleave = emb_15_interleave_0, values = (var_118_cast_fp16, var_119_cast_fp16))[name = tensor("emb_15_cast_fp16")]; + tensor var_123_begin_0 = const()[name = tensor("op_123_begin_0"), val = tensor([0, 128])]; + tensor var_123_end_0 = const()[name = tensor("op_123_end_0"), val = tensor([6, 256])]; + tensor var_123_end_mask_0 = const()[name = tensor("op_123_end_mask_0"), val = tensor([true, true])]; + tensor var_123_cast_fp16 = slice_by_index(begin = var_123_begin_0, end = var_123_end_0, end_mask = var_123_end_mask_0, x = emb_15_cast_fp16)[name = tensor("op_123_cast_fp16")]; + tensor var_125_begin_0 = const()[name = tensor("op_125_begin_0"), val = tensor([0, 0])]; + tensor var_125_end_0 = const()[name = tensor("op_125_end_0"), val = tensor([6, 128])]; + tensor var_125_end_mask_0 = const()[name = tensor("op_125_end_mask_0"), val = tensor([true, false])]; + tensor var_125_cast_fp16 = slice_by_index(begin = var_125_begin_0, end = var_125_end_0, end_mask = var_125_end_mask_0, x = emb_15_cast_fp16)[name = tensor("op_125_cast_fp16")]; + tensor time_embeds_1_interleave_0 = const()[name = tensor("time_embeds_1_interleave_0"), val = tensor(false)]; + tensor time_embeds_1_cast_fp16 = concat(axis = var_95, interleave = time_embeds_1_interleave_0, values = (var_123_cast_fp16, var_125_cast_fp16))[name = tensor("time_embeds_1_cast_fp16")]; + tensor var_133 = const()[name = tensor("op_133"), val = tensor([1, -1])]; + tensor time_embeds_cast_fp16 = reshape(shape = var_133, x = time_embeds_1_cast_fp16)[name = tensor("time_embeds_cast_fp16")]; + tensor var_136 = const()[name = tensor("op_136"), val = tensor(-1)]; + tensor add_embeds_interleave_0 = const()[name = tensor("add_embeds_interleave_0"), val = tensor(false)]; + tensor add_embeds_cast_fp16 = concat(axis = var_136, interleave = add_embeds_interleave_0, values = (text_embeds, time_embeds_cast_fp16))[name = tensor("add_embeds_cast_fp16")]; + tensor add_embedding_linear_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(1542528))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4245952))), name = tensor("add_embedding_linear_1_weight_to_fp16_palettized"), shape = tensor([1280, 2816])]; + tensor add_embedding_linear_1_bias_to_fp16 = const()[name = tensor("add_embedding_linear_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4246144)))]; + tensor linear_2_cast_fp16 = linear(bias = add_embedding_linear_1_bias_to_fp16, weight = add_embedding_linear_1_weight_to_fp16_palettized, x = add_embeds_cast_fp16)[name = tensor("linear_2_cast_fp16")]; + tensor input_11_cast_fp16 = silu(x = linear_2_cast_fp16)[name = tensor("input_11_cast_fp16")]; + tensor add_embedding_linear_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(4248768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5477632))), name = tensor("add_embedding_linear_2_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor add_embedding_linear_2_bias_to_fp16 = const()[name = tensor("add_embedding_linear_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5477824)))]; + tensor linear_3_cast_fp16 = linear(bias = add_embedding_linear_2_bias_to_fp16, weight = add_embedding_linear_2_weight_to_fp16_palettized, x = input_11_cast_fp16)[name = tensor("linear_3_cast_fp16")]; + tensor input_47_cast_fp16 = add(x = linear_1_cast_fp16, y = linear_3_cast_fp16)[name = tensor("input_47_cast_fp16")]; + tensor var_157 = const()[name = tensor("op_157"), val = tensor(1)]; + tensor var_160 = const()[name = tensor("op_160"), val = tensor([1, 1])]; + tensor var_162 = const()[name = tensor("op_162"), val = tensor([1, 1])]; + tensor sample_3_pad_type_0 = const()[name = tensor("sample_3_pad_type_0"), val = tensor("custom")]; + tensor sample_3_pad_0 = const()[name = tensor("sample_3_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor conv_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5480448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5489152))), name = tensor("conv_in_weight_to_fp16_palettized"), shape = tensor([320, 4, 3, 3])]; + tensor conv_in_bias_to_fp16 = const()[name = tensor("conv_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5489344)))]; + tensor sample_3_cast_fp16 = conv(bias = conv_in_bias_to_fp16, dilations = var_162, groups = var_157, pad = sample_3_pad_0, pad_type = sample_3_pad_type_0, strides = var_160, weight = conv_in_weight_to_fp16_palettized, x = sample)[name = tensor("sample_3_cast_fp16")]; + tensor var_166 = const()[name = tensor("op_166"), val = tensor(1)]; + tensor var_186 = const()[name = tensor("op_186"), val = tensor([1, 1])]; + tensor var_188 = const()[name = tensor("op_188"), val = tensor([1, 1])]; + tensor input_13_pad_type_0 = const()[name = tensor("input_13_pad_type_0"), val = tensor("custom")]; + tensor input_13_pad_0 = const()[name = tensor("input_13_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_conv_in_weight_to_fp16 = const()[name = tensor("controlnet_cond_embedding_conv_in_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5490048)))]; + tensor controlnet_cond_embedding_conv_in_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_conv_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5491008)))]; + tensor input_13_cast_fp16 = conv(bias = controlnet_cond_embedding_conv_in_bias_to_fp16, dilations = var_188, groups = var_166, pad = input_13_pad_0, pad_type = input_13_pad_type_0, strides = var_186, weight = controlnet_cond_embedding_conv_in_weight_to_fp16, x = controlnet_cond)[name = tensor("input_13_cast_fp16")]; + tensor input_15_cast_fp16 = silu(x = input_13_cast_fp16)[name = tensor("input_15_cast_fp16")]; + tensor var_194 = const()[name = tensor("op_194"), val = tensor([1, 1])]; + tensor var_196 = const()[name = tensor("op_196"), val = tensor([1, 1])]; + tensor input_17_pad_type_0 = const()[name = tensor("input_17_pad_type_0"), val = tensor("custom")]; + tensor input_17_pad_0 = const()[name = tensor("input_17_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5491136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5492928))), name = tensor("controlnet_cond_embedding_blocks_0_weight_to_fp16_palettized"), shape = tensor([16, 16, 3, 3])]; + tensor controlnet_cond_embedding_blocks_0_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5493120)))]; + tensor input_17_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_0_bias_to_fp16, dilations = var_196, groups = var_166, pad = input_17_pad_0, pad_type = input_17_pad_type_0, strides = var_194, weight = controlnet_cond_embedding_blocks_0_weight_to_fp16_palettized, x = input_15_cast_fp16)[name = tensor("input_17_cast_fp16")]; + tensor input_19_cast_fp16 = silu(x = input_17_cast_fp16)[name = tensor("input_19_cast_fp16")]; + tensor var_202 = const()[name = tensor("op_202"), val = tensor([2, 2])]; + tensor var_204 = const()[name = tensor("op_204"), val = tensor([1, 1])]; + tensor input_21_pad_type_0 = const()[name = tensor("input_21_pad_type_0"), val = tensor("custom")]; + tensor input_21_pad_0 = const()[name = tensor("input_21_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5493248))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5496768))), name = tensor("controlnet_cond_embedding_blocks_1_weight_to_fp16_palettized"), shape = tensor([32, 16, 3, 3])]; + tensor controlnet_cond_embedding_blocks_1_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5496960)))]; + tensor input_21_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_1_bias_to_fp16, dilations = var_204, groups = var_166, pad = input_21_pad_0, pad_type = input_21_pad_type_0, strides = var_202, weight = controlnet_cond_embedding_blocks_1_weight_to_fp16_palettized, x = input_19_cast_fp16)[name = tensor("input_21_cast_fp16")]; + tensor input_23_cast_fp16 = silu(x = input_21_cast_fp16)[name = tensor("input_23_cast_fp16")]; + tensor var_210 = const()[name = tensor("op_210"), val = tensor([1, 1])]; + tensor var_212 = const()[name = tensor("op_212"), val = tensor([1, 1])]; + tensor input_25_pad_type_0 = const()[name = tensor("input_25_pad_type_0"), val = tensor("custom")]; + tensor input_25_pad_0 = const()[name = tensor("input_25_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5497088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5504064))), name = tensor("controlnet_cond_embedding_blocks_2_weight_to_fp16_palettized"), shape = tensor([32, 32, 3, 3])]; + tensor controlnet_cond_embedding_blocks_2_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5504256)))]; + tensor input_25_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_2_bias_to_fp16, dilations = var_212, groups = var_166, pad = input_25_pad_0, pad_type = input_25_pad_type_0, strides = var_210, weight = controlnet_cond_embedding_blocks_2_weight_to_fp16_palettized, x = input_23_cast_fp16)[name = tensor("input_25_cast_fp16")]; + tensor input_27_cast_fp16 = silu(x = input_25_cast_fp16)[name = tensor("input_27_cast_fp16")]; + tensor var_218 = const()[name = tensor("op_218"), val = tensor([2, 2])]; + tensor var_220 = const()[name = tensor("op_220"), val = tensor([1, 1])]; + tensor input_29_pad_type_0 = const()[name = tensor("input_29_pad_type_0"), val = tensor("custom")]; + tensor input_29_pad_0 = const()[name = tensor("input_29_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_3_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5504384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5525184))), name = tensor("controlnet_cond_embedding_blocks_3_weight_to_fp16_palettized"), shape = tensor([96, 32, 3, 3])]; + tensor controlnet_cond_embedding_blocks_3_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5525376)))]; + tensor input_29_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_3_bias_to_fp16, dilations = var_220, groups = var_166, pad = input_29_pad_0, pad_type = input_29_pad_type_0, strides = var_218, weight = controlnet_cond_embedding_blocks_3_weight_to_fp16_palettized, x = input_27_cast_fp16)[name = tensor("input_29_cast_fp16")]; + tensor input_31_cast_fp16 = silu(x = input_29_cast_fp16)[name = tensor("input_31_cast_fp16")]; + tensor var_226 = const()[name = tensor("op_226"), val = tensor([1, 1])]; + tensor var_228 = const()[name = tensor("op_228"), val = tensor([1, 1])]; + tensor input_33_pad_type_0 = const()[name = tensor("input_33_pad_type_0"), val = tensor("custom")]; + tensor input_33_pad_0 = const()[name = tensor("input_33_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_4_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5525632))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5587904))), name = tensor("controlnet_cond_embedding_blocks_4_weight_to_fp16_palettized"), shape = tensor([96, 96, 3, 3])]; + tensor controlnet_cond_embedding_blocks_4_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_4_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5588096)))]; + tensor input_33_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_4_bias_to_fp16, dilations = var_228, groups = var_166, pad = input_33_pad_0, pad_type = input_33_pad_type_0, strides = var_226, weight = controlnet_cond_embedding_blocks_4_weight_to_fp16_palettized, x = input_31_cast_fp16)[name = tensor("input_33_cast_fp16")]; + tensor input_35_cast_fp16 = silu(x = input_33_cast_fp16)[name = tensor("input_35_cast_fp16")]; + tensor var_234 = const()[name = tensor("op_234"), val = tensor([2, 2])]; + tensor var_236 = const()[name = tensor("op_236"), val = tensor([1, 1])]; + tensor input_37_pad_type_0 = const()[name = tensor("input_37_pad_type_0"), val = tensor("custom")]; + tensor input_37_pad_0 = const()[name = tensor("input_37_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_blocks_5_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5588352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5754304))), name = tensor("controlnet_cond_embedding_blocks_5_weight_to_fp16_palettized"), shape = tensor([256, 96, 3, 3])]; + tensor controlnet_cond_embedding_blocks_5_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_blocks_5_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5754496)))]; + tensor input_37_cast_fp16 = conv(bias = controlnet_cond_embedding_blocks_5_bias_to_fp16, dilations = var_236, groups = var_166, pad = input_37_pad_0, pad_type = input_37_pad_type_0, strides = var_234, weight = controlnet_cond_embedding_blocks_5_weight_to_fp16_palettized, x = input_35_cast_fp16)[name = tensor("input_37_cast_fp16")]; + tensor input_39_cast_fp16 = silu(x = input_37_cast_fp16)[name = tensor("input_39_cast_fp16")]; + tensor var_242 = const()[name = tensor("op_242"), val = tensor([1, 1])]; + tensor var_244 = const()[name = tensor("op_244"), val = tensor([1, 1])]; + tensor controlnet_cond_pad_type_0 = const()[name = tensor("controlnet_cond_pad_type_0"), val = tensor("custom")]; + tensor controlnet_cond_pad_0 = const()[name = tensor("controlnet_cond_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor controlnet_cond_embedding_conv_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(5755072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6308096))), name = tensor("controlnet_cond_embedding_conv_out_weight_to_fp16_palettized"), shape = tensor([320, 256, 3, 3])]; + tensor controlnet_cond_embedding_conv_out_bias_to_fp16 = const()[name = tensor("controlnet_cond_embedding_conv_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6308288)))]; + tensor controlnet_cond_cast_fp16 = conv(bias = controlnet_cond_embedding_conv_out_bias_to_fp16, dilations = var_244, groups = var_166, pad = controlnet_cond_pad_0, pad_type = controlnet_cond_pad_type_0, strides = var_242, weight = controlnet_cond_embedding_conv_out_weight_to_fp16_palettized, x = input_39_cast_fp16)[name = tensor("controlnet_cond_cast_fp16")]; + tensor input_41_cast_fp16 = add(x = sample_3_cast_fp16, y = controlnet_cond_cast_fp16)[name = tensor("input_41_cast_fp16")]; + tensor var_254 = const()[name = tensor("op_254"), val = tensor(1)]; + tensor reshape_0_shape_0 = const()[name = tensor("reshape_0_shape_0"), val = tensor([1, 32, 10, 128, 128])]; + tensor reshape_0_cast_fp16 = reshape(shape = reshape_0_shape_0, x = input_41_cast_fp16)[name = tensor("reshape_0_cast_fp16")]; + tensor reduce_mean_0_axes_0 = const()[name = tensor("reduce_mean_0_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_0_keep_dims_0 = const()[name = tensor("reduce_mean_0_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_0_cast_fp16 = reduce_mean(axes = reduce_mean_0_axes_0, keep_dims = reduce_mean_0_keep_dims_0, x = reshape_0_cast_fp16)[name = tensor("reduce_mean_0_cast_fp16")]; + tensor sub_0_cast_fp16 = sub(x = reshape_0_cast_fp16, y = reduce_mean_0_cast_fp16)[name = tensor("sub_0_cast_fp16")]; + tensor square_0_cast_fp16 = square(x = sub_0_cast_fp16)[name = tensor("square_0_cast_fp16")]; + tensor reduce_mean_2_axes_0 = const()[name = tensor("reduce_mean_2_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_2_keep_dims_0 = const()[name = tensor("reduce_mean_2_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_2_cast_fp16 = reduce_mean(axes = reduce_mean_2_axes_0, keep_dims = reduce_mean_2_keep_dims_0, x = square_0_cast_fp16)[name = tensor("reduce_mean_2_cast_fp16")]; + tensor add_0_y_0_to_fp16 = const()[name = tensor("add_0_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_0_cast_fp16 = add(x = reduce_mean_2_cast_fp16, y = add_0_y_0_to_fp16)[name = tensor("add_0_cast_fp16")]; + tensor sqrt_0_cast_fp16 = sqrt(x = add_0_cast_fp16)[name = tensor("sqrt_0_cast_fp16")]; + tensor real_div_0_cast_fp16 = real_div(x = sub_0_cast_fp16, y = sqrt_0_cast_fp16)[name = tensor("real_div_0_cast_fp16")]; + tensor reshape_1_shape_0 = const()[name = tensor("reshape_1_shape_0"), val = tensor([1, 320, 128, 128])]; + tensor reshape_1_cast_fp16 = reshape(shape = reshape_1_shape_0, x = real_div_0_cast_fp16)[name = tensor("reshape_1_cast_fp16")]; + tensor add_1_mean_0_to_fp16 = const()[name = tensor("add_1_mean_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6308992)))]; + tensor add_1_variance_0_to_fp16 = const()[name = tensor("add_1_variance_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6309696)))]; + tensor add_1_gamma_0_to_fp16 = const()[name = tensor("add_1_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6310400)))]; + tensor add_1_beta_0_to_fp16 = const()[name = tensor("add_1_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6311104)))]; + tensor add_1_epsilon_0_to_fp16 = const()[name = tensor("add_1_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_1_cast_fp16 = batch_norm(beta = add_1_beta_0_to_fp16, epsilon = add_1_epsilon_0_to_fp16, gamma = add_1_gamma_0_to_fp16, mean = add_1_mean_0_to_fp16, variance = add_1_variance_0_to_fp16, x = reshape_1_cast_fp16)[name = tensor("add_1_cast_fp16")]; + tensor input_45_cast_fp16 = silu(x = add_1_cast_fp16)[name = tensor("input_45_cast_fp16")]; + tensor var_276 = const()[name = tensor("op_276"), val = tensor([1, 1])]; + tensor var_278 = const()[name = tensor("op_278"), val = tensor([1, 1])]; + tensor hidden_states_1_pad_type_0 = const()[name = tensor("hidden_states_1_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_1_pad_0 = const()[name = tensor("hidden_states_1_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_0_resnets_0_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(6311808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7003072))), name = tensor("down_blocks_0_resnets_0_conv1_weight_to_fp16_palettized"), shape = tensor([320, 320, 3, 3])]; + tensor down_blocks_0_resnets_0_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7003264)))]; + tensor hidden_states_1_cast_fp16 = conv(bias = down_blocks_0_resnets_0_conv1_bias_to_fp16, dilations = var_278, groups = var_254, pad = hidden_states_1_pad_0, pad_type = hidden_states_1_pad_type_0, strides = var_276, weight = down_blocks_0_resnets_0_conv1_weight_to_fp16_palettized, x = input_45_cast_fp16)[name = tensor("hidden_states_1_cast_fp16")]; + tensor input_49_cast_fp16 = silu(x = input_47_cast_fp16)[name = tensor("input_49_cast_fp16")]; + tensor down_blocks_0_resnets_0_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7003968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7311232))), name = tensor("down_blocks_0_resnets_0_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([320, 1280])]; + tensor down_blocks_0_resnets_0_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_0_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7311424)))]; + tensor linear_4_cast_fp16 = linear(bias = down_blocks_0_resnets_0_time_emb_proj_bias_to_fp16, weight = down_blocks_0_resnets_0_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_4_cast_fp16")]; + tensor var_287_axes_0 = const()[name = tensor("op_287_axes_0"), val = tensor([2])]; + tensor var_287_cast_fp16 = expand_dims(axes = var_287_axes_0, x = linear_4_cast_fp16)[name = tensor("op_287_cast_fp16")]; + tensor temb_1_axes_0 = const()[name = tensor("temb_1_axes_0"), val = tensor([3])]; + tensor temb_1_cast_fp16 = expand_dims(axes = temb_1_axes_0, x = var_287_cast_fp16)[name = tensor("temb_1_cast_fp16")]; + tensor input_51_cast_fp16 = add(x = hidden_states_1_cast_fp16, y = temb_1_cast_fp16)[name = tensor("input_51_cast_fp16")]; + tensor reshape_4_shape_0 = const()[name = tensor("reshape_4_shape_0"), val = tensor([1, 32, 10, 128, 128])]; + tensor reshape_4_cast_fp16 = reshape(shape = reshape_4_shape_0, x = input_51_cast_fp16)[name = tensor("reshape_4_cast_fp16")]; + tensor reduce_mean_3_axes_0 = const()[name = tensor("reduce_mean_3_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_3_keep_dims_0 = const()[name = tensor("reduce_mean_3_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_3_cast_fp16 = reduce_mean(axes = reduce_mean_3_axes_0, keep_dims = reduce_mean_3_keep_dims_0, x = reshape_4_cast_fp16)[name = tensor("reduce_mean_3_cast_fp16")]; + tensor sub_2_cast_fp16 = sub(x = reshape_4_cast_fp16, y = reduce_mean_3_cast_fp16)[name = tensor("sub_2_cast_fp16")]; + tensor square_1_cast_fp16 = square(x = sub_2_cast_fp16)[name = tensor("square_1_cast_fp16")]; + tensor reduce_mean_5_axes_0 = const()[name = tensor("reduce_mean_5_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_5_keep_dims_0 = const()[name = tensor("reduce_mean_5_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_5_cast_fp16 = reduce_mean(axes = reduce_mean_5_axes_0, keep_dims = reduce_mean_5_keep_dims_0, x = square_1_cast_fp16)[name = tensor("reduce_mean_5_cast_fp16")]; + tensor add_2_y_0_to_fp16 = const()[name = tensor("add_2_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_2_cast_fp16 = add(x = reduce_mean_5_cast_fp16, y = add_2_y_0_to_fp16)[name = tensor("add_2_cast_fp16")]; + tensor sqrt_1_cast_fp16 = sqrt(x = add_2_cast_fp16)[name = tensor("sqrt_1_cast_fp16")]; + tensor real_div_1_cast_fp16 = real_div(x = sub_2_cast_fp16, y = sqrt_1_cast_fp16)[name = tensor("real_div_1_cast_fp16")]; + tensor reshape_5_shape_0 = const()[name = tensor("reshape_5_shape_0"), val = tensor([1, 320, 128, 128])]; + tensor reshape_5_cast_fp16 = reshape(shape = reshape_5_shape_0, x = real_div_1_cast_fp16)[name = tensor("reshape_5_cast_fp16")]; + tensor add_3_gamma_0_to_fp16 = const()[name = tensor("add_3_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7312128)))]; + tensor add_3_beta_0_to_fp16 = const()[name = tensor("add_3_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7312832)))]; + tensor add_3_epsilon_0_to_fp16 = const()[name = tensor("add_3_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_3_cast_fp16 = batch_norm(beta = add_3_beta_0_to_fp16, epsilon = add_3_epsilon_0_to_fp16, gamma = add_3_gamma_0_to_fp16, mean = add_1_mean_0_to_fp16, variance = add_1_variance_0_to_fp16, x = reshape_5_cast_fp16)[name = tensor("add_3_cast_fp16")]; + tensor input_55_cast_fp16 = silu(x = add_3_cast_fp16)[name = tensor("input_55_cast_fp16")]; + tensor var_297 = const()[name = tensor("op_297"), val = tensor([1, 1])]; + tensor var_299 = const()[name = tensor("op_299"), val = tensor([1, 1])]; + tensor hidden_states_3_pad_type_0 = const()[name = tensor("hidden_states_3_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_3_pad_0 = const()[name = tensor("hidden_states_3_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_0_resnets_0_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(7313536))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8004800))), name = tensor("down_blocks_0_resnets_0_conv2_weight_to_fp16_palettized"), shape = tensor([320, 320, 3, 3])]; + tensor down_blocks_0_resnets_0_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8004992)))]; + tensor hidden_states_3_cast_fp16 = conv(bias = down_blocks_0_resnets_0_conv2_bias_to_fp16, dilations = var_299, groups = var_254, pad = hidden_states_3_pad_0, pad_type = hidden_states_3_pad_type_0, strides = var_297, weight = down_blocks_0_resnets_0_conv2_weight_to_fp16_palettized, x = input_55_cast_fp16)[name = tensor("hidden_states_3_cast_fp16")]; + tensor var_302_cast_fp16 = add(x = input_41_cast_fp16, y = hidden_states_3_cast_fp16)[name = tensor("op_302_cast_fp16")]; + tensor reshape_8_shape_0 = const()[name = tensor("reshape_8_shape_0"), val = tensor([1, 32, 10, 128, 128])]; + tensor reshape_8_cast_fp16 = reshape(shape = reshape_8_shape_0, x = var_302_cast_fp16)[name = tensor("reshape_8_cast_fp16")]; + tensor reduce_mean_6_axes_0 = const()[name = tensor("reduce_mean_6_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_6_keep_dims_0 = const()[name = tensor("reduce_mean_6_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_6_cast_fp16 = reduce_mean(axes = reduce_mean_6_axes_0, keep_dims = reduce_mean_6_keep_dims_0, x = reshape_8_cast_fp16)[name = tensor("reduce_mean_6_cast_fp16")]; + tensor sub_4_cast_fp16 = sub(x = reshape_8_cast_fp16, y = reduce_mean_6_cast_fp16)[name = tensor("sub_4_cast_fp16")]; + tensor square_2_cast_fp16 = square(x = sub_4_cast_fp16)[name = tensor("square_2_cast_fp16")]; + tensor reduce_mean_8_axes_0 = const()[name = tensor("reduce_mean_8_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_8_keep_dims_0 = const()[name = tensor("reduce_mean_8_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_8_cast_fp16 = reduce_mean(axes = reduce_mean_8_axes_0, keep_dims = reduce_mean_8_keep_dims_0, x = square_2_cast_fp16)[name = tensor("reduce_mean_8_cast_fp16")]; + tensor add_4_y_0_to_fp16 = const()[name = tensor("add_4_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_4_cast_fp16 = add(x = reduce_mean_8_cast_fp16, y = add_4_y_0_to_fp16)[name = tensor("add_4_cast_fp16")]; + tensor sqrt_2_cast_fp16 = sqrt(x = add_4_cast_fp16)[name = tensor("sqrt_2_cast_fp16")]; + tensor real_div_2_cast_fp16 = real_div(x = sub_4_cast_fp16, y = sqrt_2_cast_fp16)[name = tensor("real_div_2_cast_fp16")]; + tensor reshape_9_shape_0 = const()[name = tensor("reshape_9_shape_0"), val = tensor([1, 320, 128, 128])]; + tensor reshape_9_cast_fp16 = reshape(shape = reshape_9_shape_0, x = real_div_2_cast_fp16)[name = tensor("reshape_9_cast_fp16")]; + tensor add_5_gamma_0_to_fp16 = const()[name = tensor("add_5_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8005696)))]; + tensor add_5_beta_0_to_fp16 = const()[name = tensor("add_5_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8006400)))]; + tensor add_5_epsilon_0_to_fp16 = const()[name = tensor("add_5_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_5_cast_fp16 = batch_norm(beta = add_5_beta_0_to_fp16, epsilon = add_5_epsilon_0_to_fp16, gamma = add_5_gamma_0_to_fp16, mean = add_1_mean_0_to_fp16, variance = add_1_variance_0_to_fp16, x = reshape_9_cast_fp16)[name = tensor("add_5_cast_fp16")]; + tensor input_63_cast_fp16 = silu(x = add_5_cast_fp16)[name = tensor("input_63_cast_fp16")]; + tensor var_316 = const()[name = tensor("op_316"), val = tensor([1, 1])]; + tensor var_318 = const()[name = tensor("op_318"), val = tensor([1, 1])]; + tensor hidden_states_5_pad_type_0 = const()[name = tensor("hidden_states_5_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_5_pad_0 = const()[name = tensor("hidden_states_5_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_0_resnets_1_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8007104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8698368))), name = tensor("down_blocks_0_resnets_1_conv1_weight_to_fp16_palettized"), shape = tensor([320, 320, 3, 3])]; + tensor down_blocks_0_resnets_1_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8698560)))]; + tensor hidden_states_5_cast_fp16 = conv(bias = down_blocks_0_resnets_1_conv1_bias_to_fp16, dilations = var_318, groups = var_254, pad = hidden_states_5_pad_0, pad_type = hidden_states_5_pad_type_0, strides = var_316, weight = down_blocks_0_resnets_1_conv1_weight_to_fp16_palettized, x = input_63_cast_fp16)[name = tensor("hidden_states_5_cast_fp16")]; + tensor down_blocks_0_resnets_1_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(8699264))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9006528))), name = tensor("down_blocks_0_resnets_1_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([320, 1280])]; + tensor down_blocks_0_resnets_1_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_1_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9006720)))]; + tensor linear_5_cast_fp16 = linear(bias = down_blocks_0_resnets_1_time_emb_proj_bias_to_fp16, weight = down_blocks_0_resnets_1_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_5_cast_fp16")]; + tensor var_327_axes_0 = const()[name = tensor("op_327_axes_0"), val = tensor([2])]; + tensor var_327_cast_fp16 = expand_dims(axes = var_327_axes_0, x = linear_5_cast_fp16)[name = tensor("op_327_cast_fp16")]; + tensor temb_3_axes_0 = const()[name = tensor("temb_3_axes_0"), val = tensor([3])]; + tensor temb_3_cast_fp16 = expand_dims(axes = temb_3_axes_0, x = var_327_cast_fp16)[name = tensor("temb_3_cast_fp16")]; + tensor input_67_cast_fp16 = add(x = hidden_states_5_cast_fp16, y = temb_3_cast_fp16)[name = tensor("input_67_cast_fp16")]; + tensor reshape_12_shape_0 = const()[name = tensor("reshape_12_shape_0"), val = tensor([1, 32, 10, 128, 128])]; + tensor reshape_12_cast_fp16 = reshape(shape = reshape_12_shape_0, x = input_67_cast_fp16)[name = tensor("reshape_12_cast_fp16")]; + tensor reduce_mean_9_axes_0 = const()[name = tensor("reduce_mean_9_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_9_keep_dims_0 = const()[name = tensor("reduce_mean_9_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_9_cast_fp16 = reduce_mean(axes = reduce_mean_9_axes_0, keep_dims = reduce_mean_9_keep_dims_0, x = reshape_12_cast_fp16)[name = tensor("reduce_mean_9_cast_fp16")]; + tensor sub_6_cast_fp16 = sub(x = reshape_12_cast_fp16, y = reduce_mean_9_cast_fp16)[name = tensor("sub_6_cast_fp16")]; + tensor square_3_cast_fp16 = square(x = sub_6_cast_fp16)[name = tensor("square_3_cast_fp16")]; + tensor reduce_mean_11_axes_0 = const()[name = tensor("reduce_mean_11_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_11_keep_dims_0 = const()[name = tensor("reduce_mean_11_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_11_cast_fp16 = reduce_mean(axes = reduce_mean_11_axes_0, keep_dims = reduce_mean_11_keep_dims_0, x = square_3_cast_fp16)[name = tensor("reduce_mean_11_cast_fp16")]; + tensor add_6_y_0_to_fp16 = const()[name = tensor("add_6_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_6_cast_fp16 = add(x = reduce_mean_11_cast_fp16, y = add_6_y_0_to_fp16)[name = tensor("add_6_cast_fp16")]; + tensor sqrt_3_cast_fp16 = sqrt(x = add_6_cast_fp16)[name = tensor("sqrt_3_cast_fp16")]; + tensor real_div_3_cast_fp16 = real_div(x = sub_6_cast_fp16, y = sqrt_3_cast_fp16)[name = tensor("real_div_3_cast_fp16")]; + tensor reshape_13_shape_0 = const()[name = tensor("reshape_13_shape_0"), val = tensor([1, 320, 128, 128])]; + tensor reshape_13_cast_fp16 = reshape(shape = reshape_13_shape_0, x = real_div_3_cast_fp16)[name = tensor("reshape_13_cast_fp16")]; + tensor add_7_gamma_0_to_fp16 = const()[name = tensor("add_7_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9007424)))]; + tensor add_7_beta_0_to_fp16 = const()[name = tensor("add_7_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9008128)))]; + tensor add_7_epsilon_0_to_fp16 = const()[name = tensor("add_7_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_7_cast_fp16 = batch_norm(beta = add_7_beta_0_to_fp16, epsilon = add_7_epsilon_0_to_fp16, gamma = add_7_gamma_0_to_fp16, mean = add_1_mean_0_to_fp16, variance = add_1_variance_0_to_fp16, x = reshape_13_cast_fp16)[name = tensor("add_7_cast_fp16")]; + tensor input_71_cast_fp16 = silu(x = add_7_cast_fp16)[name = tensor("input_71_cast_fp16")]; + tensor var_337 = const()[name = tensor("op_337"), val = tensor([1, 1])]; + tensor var_339 = const()[name = tensor("op_339"), val = tensor([1, 1])]; + tensor hidden_states_7_pad_type_0 = const()[name = tensor("hidden_states_7_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_7_pad_0 = const()[name = tensor("hidden_states_7_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_0_resnets_1_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9008832))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9700096))), name = tensor("down_blocks_0_resnets_1_conv2_weight_to_fp16_palettized"), shape = tensor([320, 320, 3, 3])]; + tensor down_blocks_0_resnets_1_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_0_resnets_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9700288)))]; + tensor hidden_states_7_cast_fp16 = conv(bias = down_blocks_0_resnets_1_conv2_bias_to_fp16, dilations = var_339, groups = var_254, pad = hidden_states_7_pad_0, pad_type = hidden_states_7_pad_type_0, strides = var_337, weight = down_blocks_0_resnets_1_conv2_weight_to_fp16_palettized, x = input_71_cast_fp16)[name = tensor("hidden_states_7_cast_fp16")]; + tensor var_342_cast_fp16 = add(x = var_302_cast_fp16, y = hidden_states_7_cast_fp16)[name = tensor("op_342_cast_fp16")]; + tensor var_348 = const()[name = tensor("op_348"), val = tensor([2, 2])]; + tensor var_350 = const()[name = tensor("op_350"), val = tensor([1, 1])]; + tensor input_75_pad_type_0 = const()[name = tensor("input_75_pad_type_0"), val = tensor("custom")]; + tensor input_75_pad_0 = const()[name = tensor("input_75_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_0_downsamplers_0_conv_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(9700992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10392256))), name = tensor("down_blocks_0_downsamplers_0_conv_weight_to_fp16_palettized"), shape = tensor([320, 320, 3, 3])]; + tensor down_blocks_0_downsamplers_0_conv_bias_to_fp16 = const()[name = tensor("down_blocks_0_downsamplers_0_conv_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10392448)))]; + tensor input_75_cast_fp16 = conv(bias = down_blocks_0_downsamplers_0_conv_bias_to_fp16, dilations = var_350, groups = var_254, pad = input_75_pad_0, pad_type = input_75_pad_type_0, strides = var_348, weight = down_blocks_0_downsamplers_0_conv_weight_to_fp16_palettized, x = var_342_cast_fp16)[name = tensor("input_75_cast_fp16")]; + tensor var_370 = const()[name = tensor("op_370"), val = tensor(1)]; + tensor reshape_16_shape_0 = const()[name = tensor("reshape_16_shape_0"), val = tensor([1, 32, 10, 64, 64])]; + tensor reshape_16_cast_fp16 = reshape(shape = reshape_16_shape_0, x = input_75_cast_fp16)[name = tensor("reshape_16_cast_fp16")]; + tensor reduce_mean_12_axes_0 = const()[name = tensor("reduce_mean_12_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_12_keep_dims_0 = const()[name = tensor("reduce_mean_12_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_12_cast_fp16 = reduce_mean(axes = reduce_mean_12_axes_0, keep_dims = reduce_mean_12_keep_dims_0, x = reshape_16_cast_fp16)[name = tensor("reduce_mean_12_cast_fp16")]; + tensor sub_8_cast_fp16 = sub(x = reshape_16_cast_fp16, y = reduce_mean_12_cast_fp16)[name = tensor("sub_8_cast_fp16")]; + tensor square_4_cast_fp16 = square(x = sub_8_cast_fp16)[name = tensor("square_4_cast_fp16")]; + tensor reduce_mean_14_axes_0 = const()[name = tensor("reduce_mean_14_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_14_keep_dims_0 = const()[name = tensor("reduce_mean_14_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_14_cast_fp16 = reduce_mean(axes = reduce_mean_14_axes_0, keep_dims = reduce_mean_14_keep_dims_0, x = square_4_cast_fp16)[name = tensor("reduce_mean_14_cast_fp16")]; + tensor add_8_y_0_to_fp16 = const()[name = tensor("add_8_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_8_cast_fp16 = add(x = reduce_mean_14_cast_fp16, y = add_8_y_0_to_fp16)[name = tensor("add_8_cast_fp16")]; + tensor sqrt_4_cast_fp16 = sqrt(x = add_8_cast_fp16)[name = tensor("sqrt_4_cast_fp16")]; + tensor real_div_4_cast_fp16 = real_div(x = sub_8_cast_fp16, y = sqrt_4_cast_fp16)[name = tensor("real_div_4_cast_fp16")]; + tensor reshape_17_shape_0 = const()[name = tensor("reshape_17_shape_0"), val = tensor([1, 320, 64, 64])]; + tensor reshape_17_cast_fp16 = reshape(shape = reshape_17_shape_0, x = real_div_4_cast_fp16)[name = tensor("reshape_17_cast_fp16")]; + tensor add_9_gamma_0_to_fp16 = const()[name = tensor("add_9_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10393152)))]; + tensor add_9_beta_0_to_fp16 = const()[name = tensor("add_9_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10393856)))]; + tensor add_9_epsilon_0_to_fp16 = const()[name = tensor("add_9_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_9_cast_fp16 = batch_norm(beta = add_9_beta_0_to_fp16, epsilon = add_9_epsilon_0_to_fp16, gamma = add_9_gamma_0_to_fp16, mean = add_1_mean_0_to_fp16, variance = add_1_variance_0_to_fp16, x = reshape_17_cast_fp16)[name = tensor("add_9_cast_fp16")]; + tensor input_79_cast_fp16 = silu(x = add_9_cast_fp16)[name = tensor("input_79_cast_fp16")]; + tensor var_397 = const()[name = tensor("op_397"), val = tensor([1, 1])]; + tensor var_399 = const()[name = tensor("op_399"), val = tensor([1, 1])]; + tensor hidden_states_11_pad_type_0 = const()[name = tensor("hidden_states_11_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_11_pad_0 = const()[name = tensor("hidden_states_11_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_1_resnets_0_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(10394560))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11777024))), name = tensor("down_blocks_1_resnets_0_conv1_weight_to_fp16_palettized"), shape = tensor([640, 320, 3, 3])]; + tensor down_blocks_1_resnets_0_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11777216)))]; + tensor hidden_states_11_cast_fp16 = conv(bias = down_blocks_1_resnets_0_conv1_bias_to_fp16, dilations = var_399, groups = var_370, pad = hidden_states_11_pad_0, pad_type = hidden_states_11_pad_type_0, strides = var_397, weight = down_blocks_1_resnets_0_conv1_weight_to_fp16_palettized, x = input_79_cast_fp16)[name = tensor("hidden_states_11_cast_fp16")]; + tensor down_blocks_1_resnets_0_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(11778560))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12393024))), name = tensor("down_blocks_1_resnets_0_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([640, 1280])]; + tensor down_blocks_1_resnets_0_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_0_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12393216)))]; + tensor linear_6_cast_fp16 = linear(bias = down_blocks_1_resnets_0_time_emb_proj_bias_to_fp16, weight = down_blocks_1_resnets_0_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_6_cast_fp16")]; + tensor var_408_axes_0 = const()[name = tensor("op_408_axes_0"), val = tensor([2])]; + tensor var_408_cast_fp16 = expand_dims(axes = var_408_axes_0, x = linear_6_cast_fp16)[name = tensor("op_408_cast_fp16")]; + tensor temb_5_axes_0 = const()[name = tensor("temb_5_axes_0"), val = tensor([3])]; + tensor temb_5_cast_fp16 = expand_dims(axes = temb_5_axes_0, x = var_408_cast_fp16)[name = tensor("temb_5_cast_fp16")]; + tensor input_83_cast_fp16 = add(x = hidden_states_11_cast_fp16, y = temb_5_cast_fp16)[name = tensor("input_83_cast_fp16")]; + tensor reshape_20_shape_0 = const()[name = tensor("reshape_20_shape_0"), val = tensor([1, 32, 20, 64, 64])]; + tensor reshape_20_cast_fp16 = reshape(shape = reshape_20_shape_0, x = input_83_cast_fp16)[name = tensor("reshape_20_cast_fp16")]; + tensor reduce_mean_15_axes_0 = const()[name = tensor("reduce_mean_15_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_15_keep_dims_0 = const()[name = tensor("reduce_mean_15_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_15_cast_fp16 = reduce_mean(axes = reduce_mean_15_axes_0, keep_dims = reduce_mean_15_keep_dims_0, x = reshape_20_cast_fp16)[name = tensor("reduce_mean_15_cast_fp16")]; + tensor sub_10_cast_fp16 = sub(x = reshape_20_cast_fp16, y = reduce_mean_15_cast_fp16)[name = tensor("sub_10_cast_fp16")]; + tensor square_5_cast_fp16 = square(x = sub_10_cast_fp16)[name = tensor("square_5_cast_fp16")]; + tensor reduce_mean_17_axes_0 = const()[name = tensor("reduce_mean_17_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_17_keep_dims_0 = const()[name = tensor("reduce_mean_17_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_17_cast_fp16 = reduce_mean(axes = reduce_mean_17_axes_0, keep_dims = reduce_mean_17_keep_dims_0, x = square_5_cast_fp16)[name = tensor("reduce_mean_17_cast_fp16")]; + tensor add_10_y_0_to_fp16 = const()[name = tensor("add_10_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_10_cast_fp16 = add(x = reduce_mean_17_cast_fp16, y = add_10_y_0_to_fp16)[name = tensor("add_10_cast_fp16")]; + tensor sqrt_5_cast_fp16 = sqrt(x = add_10_cast_fp16)[name = tensor("sqrt_5_cast_fp16")]; + tensor real_div_5_cast_fp16 = real_div(x = sub_10_cast_fp16, y = sqrt_5_cast_fp16)[name = tensor("real_div_5_cast_fp16")]; + tensor reshape_21_shape_0 = const()[name = tensor("reshape_21_shape_0"), val = tensor([1, 640, 64, 64])]; + tensor reshape_21_cast_fp16 = reshape(shape = reshape_21_shape_0, x = real_div_5_cast_fp16)[name = tensor("reshape_21_cast_fp16")]; + tensor add_11_mean_0_to_fp16 = const()[name = tensor("add_11_mean_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12394560)))]; + tensor add_11_variance_0_to_fp16 = const()[name = tensor("add_11_variance_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12395904)))]; + tensor add_11_gamma_0_to_fp16 = const()[name = tensor("add_11_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12397248)))]; + tensor add_11_beta_0_to_fp16 = const()[name = tensor("add_11_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12398592)))]; + tensor add_11_epsilon_0_to_fp16 = const()[name = tensor("add_11_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_11_cast_fp16 = batch_norm(beta = add_11_beta_0_to_fp16, epsilon = add_11_epsilon_0_to_fp16, gamma = add_11_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_21_cast_fp16)[name = tensor("add_11_cast_fp16")]; + tensor input_87_cast_fp16 = silu(x = add_11_cast_fp16)[name = tensor("input_87_cast_fp16")]; + tensor var_418 = const()[name = tensor("op_418"), val = tensor([1, 1])]; + tensor var_420 = const()[name = tensor("op_420"), val = tensor([1, 1])]; + tensor hidden_states_13_pad_type_0 = const()[name = tensor("hidden_states_13_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_13_pad_0 = const()[name = tensor("hidden_states_13_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_1_resnets_0_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(12399936))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15164800))), name = tensor("down_blocks_1_resnets_0_conv2_weight_to_fp16_palettized"), shape = tensor([640, 640, 3, 3])]; + tensor down_blocks_1_resnets_0_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15164992)))]; + tensor hidden_states_13_cast_fp16 = conv(bias = down_blocks_1_resnets_0_conv2_bias_to_fp16, dilations = var_420, groups = var_370, pad = hidden_states_13_pad_0, pad_type = hidden_states_13_pad_type_0, strides = var_418, weight = down_blocks_1_resnets_0_conv2_weight_to_fp16_palettized, x = input_87_cast_fp16)[name = tensor("hidden_states_13_cast_fp16")]; + tensor var_425 = const()[name = tensor("op_425"), val = tensor([1, 1])]; + tensor var_427 = const()[name = tensor("op_427"), val = tensor([1, 1])]; + tensor input_tensor_1_pad_type_0 = const()[name = tensor("input_tensor_1_pad_type_0"), val = tensor("custom")]; + tensor input_tensor_1_pad_0 = const()[name = tensor("input_tensor_1_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor down_blocks_1_resnets_0_conv_shortcut_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15166336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15320000))), name = tensor("down_blocks_1_resnets_0_conv_shortcut_weight_to_fp16_palettized"), shape = tensor([640, 320, 1, 1])]; + tensor down_blocks_1_resnets_0_conv_shortcut_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_0_conv_shortcut_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15320192)))]; + tensor input_tensor_1_cast_fp16 = conv(bias = down_blocks_1_resnets_0_conv_shortcut_bias_to_fp16, dilations = var_427, groups = var_370, pad = input_tensor_1_pad_0, pad_type = input_tensor_1_pad_type_0, strides = var_425, weight = down_blocks_1_resnets_0_conv_shortcut_weight_to_fp16_palettized, x = input_75_cast_fp16)[name = tensor("input_tensor_1_cast_fp16")]; + tensor var_430_cast_fp16 = add(x = input_tensor_1_cast_fp16, y = hidden_states_13_cast_fp16)[name = tensor("op_430_cast_fp16")]; + tensor reshape_24_shape_0 = const()[name = tensor("reshape_24_shape_0"), val = tensor([1, 32, 20, 64, 64])]; + tensor reshape_24_cast_fp16 = reshape(shape = reshape_24_shape_0, x = var_430_cast_fp16)[name = tensor("reshape_24_cast_fp16")]; + tensor reduce_mean_18_axes_0 = const()[name = tensor("reduce_mean_18_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_18_keep_dims_0 = const()[name = tensor("reduce_mean_18_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_18_cast_fp16 = reduce_mean(axes = reduce_mean_18_axes_0, keep_dims = reduce_mean_18_keep_dims_0, x = reshape_24_cast_fp16)[name = tensor("reduce_mean_18_cast_fp16")]; + tensor sub_12_cast_fp16 = sub(x = reshape_24_cast_fp16, y = reduce_mean_18_cast_fp16)[name = tensor("sub_12_cast_fp16")]; + tensor square_6_cast_fp16 = square(x = sub_12_cast_fp16)[name = tensor("square_6_cast_fp16")]; + tensor reduce_mean_20_axes_0 = const()[name = tensor("reduce_mean_20_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_20_keep_dims_0 = const()[name = tensor("reduce_mean_20_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_20_cast_fp16 = reduce_mean(axes = reduce_mean_20_axes_0, keep_dims = reduce_mean_20_keep_dims_0, x = square_6_cast_fp16)[name = tensor("reduce_mean_20_cast_fp16")]; + tensor add_12_y_0_to_fp16 = const()[name = tensor("add_12_y_0_to_fp16"), val = tensor(0x1.1p-20)]; + tensor add_12_cast_fp16 = add(x = reduce_mean_20_cast_fp16, y = add_12_y_0_to_fp16)[name = tensor("add_12_cast_fp16")]; + tensor sqrt_6_cast_fp16 = sqrt(x = add_12_cast_fp16)[name = tensor("sqrt_6_cast_fp16")]; + tensor real_div_6_cast_fp16 = real_div(x = sub_12_cast_fp16, y = sqrt_6_cast_fp16)[name = tensor("real_div_6_cast_fp16")]; + tensor reshape_25_shape_0 = const()[name = tensor("reshape_25_shape_0"), val = tensor([1, 640, 64, 64])]; + tensor reshape_25_cast_fp16 = reshape(shape = reshape_25_shape_0, x = real_div_6_cast_fp16)[name = tensor("reshape_25_cast_fp16")]; + tensor add_13_gamma_0_to_fp16 = const()[name = tensor("add_13_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15321536)))]; + tensor add_13_beta_0_to_fp16 = const()[name = tensor("add_13_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15322880)))]; + tensor add_13_epsilon_0_to_fp16 = const()[name = tensor("add_13_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_13_cast_fp16 = batch_norm(beta = add_13_beta_0_to_fp16, epsilon = add_13_epsilon_0_to_fp16, gamma = add_13_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_25_cast_fp16)[name = tensor("add_13_cast_fp16")]; + tensor var_452 = const()[name = tensor("op_452"), val = tensor([0, 2, 3, 1])]; + tensor var_456 = const()[name = tensor("op_456"), val = tensor([1, 4096, 640])]; + tensor transpose_689 = transpose(perm = var_452, x = add_13_cast_fp16)[name = tensor("transpose_689")]; + tensor input_91_cast_fp16 = reshape(shape = var_456, x = transpose_689)[name = tensor("input_91_cast_fp16")]; + tensor down_blocks_1_attentions_0_proj_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15324224))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15631488))), name = tensor("down_blocks_1_attentions_0_proj_in_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_proj_in_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_proj_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15631680)))]; + tensor linear_7_cast_fp16 = linear(bias = down_blocks_1_attentions_0_proj_in_bias_to_fp16, weight = down_blocks_1_attentions_0_proj_in_weight_to_fp16_palettized, x = input_91_cast_fp16)[name = tensor("linear_7_cast_fp16")]; + tensor hidden_states_21_axes_0 = const()[name = tensor("hidden_states_21_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15633024)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15634368)))]; + tensor var_366_to_fp16 = const()[name = tensor("op_366_to_fp16"), val = tensor(0x1.5p-17)]; + tensor hidden_states_21_cast_fp16 = layer_norm(axes = hidden_states_21_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_0_norm1_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_0_norm1_weight_to_fp16, x = linear_7_cast_fp16)[name = tensor("hidden_states_21_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15635712))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15942976))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_8_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_8_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(15943168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16250432))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_9_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_9_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16250624))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16557888))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_10_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_21_cast_fp16)[name = tensor("linear_10_cast_fp16")]; + tensor var_489 = const()[name = tensor("op_489"), val = tensor([1, -1, 10, 64])]; + tensor var_490_cast_fp16 = reshape(shape = var_489, x = linear_8_cast_fp16)[name = tensor("op_490_cast_fp16")]; + tensor var_492 = const()[name = tensor("op_492"), val = tensor([1, -1, 10, 64])]; + tensor var_493_cast_fp16 = reshape(shape = var_492, x = linear_9_cast_fp16)[name = tensor("op_493_cast_fp16")]; + tensor var_495 = const()[name = tensor("op_495"), val = tensor([1, -1, 10, 64])]; + tensor var_496_cast_fp16 = reshape(shape = var_495, x = linear_10_cast_fp16)[name = tensor("op_496_cast_fp16")]; + tensor value_3_perm_0 = const()[name = tensor("value_3_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_7_y_0_to_fp16 = const()[name = tensor("mul_7_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_7_cast_fp16 = mul(x = var_490_cast_fp16, y = mul_7_y_0_to_fp16)[name = tensor("mul_7_cast_fp16")]; + tensor matmul_0_transpose_y_0 = const()[name = tensor("matmul_0_transpose_y_0"), val = tensor(true)]; + tensor matmul_0_transpose_x_0 = const()[name = tensor("matmul_0_transpose_x_0"), val = tensor(false)]; + tensor transpose_272_perm_0 = const()[name = tensor("transpose_272_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_273_perm_0 = const()[name = tensor("transpose_273_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_687 = transpose(perm = transpose_273_perm_0, x = var_493_cast_fp16)[name = tensor("transpose_687")]; + tensor transpose_688 = transpose(perm = transpose_272_perm_0, x = mul_7_cast_fp16)[name = tensor("transpose_688")]; + tensor matmul_0_cast_fp16 = matmul(transpose_x = matmul_0_transpose_x_0, transpose_y = matmul_0_transpose_y_0, x = transpose_688, y = transpose_687)[name = tensor("matmul_0_cast_fp16")]; + tensor softmax_0_axis_0 = const()[name = tensor("softmax_0_axis_0"), val = tensor(-1)]; + tensor softmax_0_cast_fp16 = softmax(axis = softmax_0_axis_0, x = matmul_0_cast_fp16)[name = tensor("softmax_0_cast_fp16")]; + tensor hidden_states_23_transpose_x_0 = const()[name = tensor("hidden_states_23_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_23_transpose_y_0 = const()[name = tensor("hidden_states_23_transpose_y_0"), val = tensor(false)]; + tensor transpose_686 = transpose(perm = value_3_perm_0, x = var_496_cast_fp16)[name = tensor("transpose_686")]; + tensor hidden_states_23_cast_fp16 = matmul(transpose_x = hidden_states_23_transpose_x_0, transpose_y = hidden_states_23_transpose_y_0, x = softmax_0_cast_fp16, y = transpose_686)[name = tensor("hidden_states_23_cast_fp16")]; + tensor var_499_perm_0 = const()[name = tensor("op_499_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_503 = const()[name = tensor("op_503"), val = tensor([1, -1, 640])]; + tensor transpose_685 = transpose(perm = var_499_perm_0, x = hidden_states_23_cast_fp16)[name = tensor("transpose_685")]; + tensor hidden_states_25_cast_fp16 = reshape(shape = var_503, x = transpose_685)[name = tensor("hidden_states_25_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16558080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16865344))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16865536)))]; + tensor linear_11_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_25_cast_fp16)[name = tensor("linear_11_cast_fp16")]; + tensor input_97_cast_fp16 = add(x = linear_11_cast_fp16, y = linear_7_cast_fp16)[name = tensor("input_97_cast_fp16")]; + tensor input_99_axes_0 = const()[name = tensor("input_99_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16866880)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16868224)))]; + tensor input_99_cast_fp16 = layer_norm(axes = input_99_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_0_norm2_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_0_norm2_weight_to_fp16, x = input_97_cast_fp16)[name = tensor("input_99_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(16869568))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17176832))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_12_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized, x = input_99_cast_fp16)[name = tensor("linear_12_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(17177024))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18160128))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_13_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_13_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(18160320))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19143424))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_14_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_14_cast_fp16")]; + tensor var_535 = const()[name = tensor("op_535"), val = tensor([1, -1, 10, 64])]; + tensor var_536_cast_fp16 = reshape(shape = var_535, x = linear_12_cast_fp16)[name = tensor("op_536_cast_fp16")]; + tensor var_538 = const()[name = tensor("op_538"), val = tensor([1, -1, 10, 64])]; + tensor var_539_cast_fp16 = reshape(shape = var_538, x = linear_13_cast_fp16)[name = tensor("op_539_cast_fp16")]; + tensor var_541 = const()[name = tensor("op_541"), val = tensor([1, -1, 10, 64])]; + tensor var_542_cast_fp16 = reshape(shape = var_541, x = linear_14_cast_fp16)[name = tensor("op_542_cast_fp16")]; + tensor value_7_perm_0 = const()[name = tensor("value_7_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_8_y_0_to_fp16 = const()[name = tensor("mul_8_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_8_cast_fp16 = mul(x = var_536_cast_fp16, y = mul_8_y_0_to_fp16)[name = tensor("mul_8_cast_fp16")]; + tensor matmul_1_transpose_y_0 = const()[name = tensor("matmul_1_transpose_y_0"), val = tensor(true)]; + tensor matmul_1_transpose_x_0 = const()[name = tensor("matmul_1_transpose_x_0"), val = tensor(false)]; + tensor transpose_274_perm_0 = const()[name = tensor("transpose_274_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_275_perm_0 = const()[name = tensor("transpose_275_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_683 = transpose(perm = transpose_275_perm_0, x = var_539_cast_fp16)[name = tensor("transpose_683")]; + tensor transpose_684 = transpose(perm = transpose_274_perm_0, x = mul_8_cast_fp16)[name = tensor("transpose_684")]; + tensor matmul_1_cast_fp16 = matmul(transpose_x = matmul_1_transpose_x_0, transpose_y = matmul_1_transpose_y_0, x = transpose_684, y = transpose_683)[name = tensor("matmul_1_cast_fp16")]; + tensor softmax_1_axis_0 = const()[name = tensor("softmax_1_axis_0"), val = tensor(-1)]; + tensor softmax_1_cast_fp16 = softmax(axis = softmax_1_axis_0, x = matmul_1_cast_fp16)[name = tensor("softmax_1_cast_fp16")]; + tensor hidden_states_29_transpose_x_0 = const()[name = tensor("hidden_states_29_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_29_transpose_y_0 = const()[name = tensor("hidden_states_29_transpose_y_0"), val = tensor(false)]; + tensor transpose_682 = transpose(perm = value_7_perm_0, x = var_542_cast_fp16)[name = tensor("transpose_682")]; + tensor hidden_states_29_cast_fp16 = matmul(transpose_x = hidden_states_29_transpose_x_0, transpose_y = hidden_states_29_transpose_y_0, x = softmax_1_cast_fp16, y = transpose_682)[name = tensor("hidden_states_29_cast_fp16")]; + tensor var_545_perm_0 = const()[name = tensor("op_545_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_549 = const()[name = tensor("op_549"), val = tensor([1, -1, 640])]; + tensor transpose_681 = transpose(perm = var_545_perm_0, x = hidden_states_29_cast_fp16)[name = tensor("transpose_681")]; + tensor hidden_states_31_cast_fp16 = reshape(shape = var_549, x = transpose_681)[name = tensor("hidden_states_31_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19143616))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19450880))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19451072)))]; + tensor linear_15_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_31_cast_fp16)[name = tensor("linear_15_cast_fp16")]; + tensor input_105_cast_fp16 = add(x = linear_15_cast_fp16, y = input_97_cast_fp16)[name = tensor("input_105_cast_fp16")]; + tensor input_107_axes_0 = const()[name = tensor("input_107_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19452416)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19453760)))]; + tensor input_107_cast_fp16 = layer_norm(axes = input_107_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_0_norm3_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_0_norm3_weight_to_fp16, x = input_105_cast_fp16)[name = tensor("input_107_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(19455104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21912768))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([5120, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21912960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21916864))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([5120])]; + tensor linear_16_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_1_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized, x = input_107_cast_fp16)[name = tensor("linear_16_cast_fp16")]; + tensor var_571_split_sizes_0 = const()[name = tensor("op_571_split_sizes_0"), val = tensor([2560, 2560])]; + tensor var_571_axis_0 = const()[name = tensor("op_571_axis_0"), val = tensor(-1)]; + tensor var_571_cast_fp16_0, tensor var_571_cast_fp16_1 = split(axis = var_571_axis_0, split_sizes = var_571_split_sizes_0, x = linear_16_cast_fp16)[name = tensor("op_571_cast_fp16")]; + tensor var_573_mode_0 = const()[name = tensor("op_573_mode_0"), val = tensor("EXACT")]; + tensor var_573_cast_fp16 = gelu(mode = var_573_mode_0, x = var_571_cast_fp16_1)[name = tensor("op_573_cast_fp16")]; + tensor input_109_cast_fp16 = mul(x = var_571_cast_fp16_0, y = var_573_cast_fp16)[name = tensor("input_109_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(21917056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23145920))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized"), shape = tensor([640, 2560])]; + tensor down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23146112)))]; + tensor linear_17_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized, x = input_109_cast_fp16)[name = tensor("linear_17_cast_fp16")]; + tensor hidden_states_39_cast_fp16 = add(x = linear_17_cast_fp16, y = input_105_cast_fp16)[name = tensor("hidden_states_39_cast_fp16")]; + tensor hidden_states_41_axes_0 = const()[name = tensor("hidden_states_41_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23147456)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23148800)))]; + tensor hidden_states_41_cast_fp16 = layer_norm(axes = hidden_states_41_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_1_norm1_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_1_norm1_weight_to_fp16, x = hidden_states_39_cast_fp16)[name = tensor("hidden_states_41_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23150144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23457408))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_18_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_41_cast_fp16)[name = tensor("linear_18_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23457600))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23764864))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_19_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_41_cast_fp16)[name = tensor("linear_19_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(23765056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24072320))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_20_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_41_cast_fp16)[name = tensor("linear_20_cast_fp16")]; + tensor var_608 = const()[name = tensor("op_608"), val = tensor([1, -1, 10, 64])]; + tensor var_609_cast_fp16 = reshape(shape = var_608, x = linear_18_cast_fp16)[name = tensor("op_609_cast_fp16")]; + tensor var_611 = const()[name = tensor("op_611"), val = tensor([1, -1, 10, 64])]; + tensor var_612_cast_fp16 = reshape(shape = var_611, x = linear_19_cast_fp16)[name = tensor("op_612_cast_fp16")]; + tensor var_614 = const()[name = tensor("op_614"), val = tensor([1, -1, 10, 64])]; + tensor var_615_cast_fp16 = reshape(shape = var_614, x = linear_20_cast_fp16)[name = tensor("op_615_cast_fp16")]; + tensor value_11_perm_0 = const()[name = tensor("value_11_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_9_y_0_to_fp16 = const()[name = tensor("mul_9_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_9_cast_fp16 = mul(x = var_609_cast_fp16, y = mul_9_y_0_to_fp16)[name = tensor("mul_9_cast_fp16")]; + tensor matmul_2_transpose_y_0 = const()[name = tensor("matmul_2_transpose_y_0"), val = tensor(true)]; + tensor matmul_2_transpose_x_0 = const()[name = tensor("matmul_2_transpose_x_0"), val = tensor(false)]; + tensor transpose_276_perm_0 = const()[name = tensor("transpose_276_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_277_perm_0 = const()[name = tensor("transpose_277_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_679 = transpose(perm = transpose_277_perm_0, x = var_612_cast_fp16)[name = tensor("transpose_679")]; + tensor transpose_680 = transpose(perm = transpose_276_perm_0, x = mul_9_cast_fp16)[name = tensor("transpose_680")]; + tensor matmul_2_cast_fp16 = matmul(transpose_x = matmul_2_transpose_x_0, transpose_y = matmul_2_transpose_y_0, x = transpose_680, y = transpose_679)[name = tensor("matmul_2_cast_fp16")]; + tensor softmax_2_axis_0 = const()[name = tensor("softmax_2_axis_0"), val = tensor(-1)]; + tensor softmax_2_cast_fp16 = softmax(axis = softmax_2_axis_0, x = matmul_2_cast_fp16)[name = tensor("softmax_2_cast_fp16")]; + tensor hidden_states_43_transpose_x_0 = const()[name = tensor("hidden_states_43_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_43_transpose_y_0 = const()[name = tensor("hidden_states_43_transpose_y_0"), val = tensor(false)]; + tensor transpose_678 = transpose(perm = value_11_perm_0, x = var_615_cast_fp16)[name = tensor("transpose_678")]; + tensor hidden_states_43_cast_fp16 = matmul(transpose_x = hidden_states_43_transpose_x_0, transpose_y = hidden_states_43_transpose_y_0, x = softmax_2_cast_fp16, y = transpose_678)[name = tensor("hidden_states_43_cast_fp16")]; + tensor var_618_perm_0 = const()[name = tensor("op_618_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_622 = const()[name = tensor("op_622"), val = tensor([1, -1, 640])]; + tensor transpose_677 = transpose(perm = var_618_perm_0, x = hidden_states_43_cast_fp16)[name = tensor("transpose_677")]; + tensor hidden_states_45_cast_fp16 = reshape(shape = var_622, x = transpose_677)[name = tensor("hidden_states_45_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24072512))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24379776))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24379968)))]; + tensor linear_21_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_45_cast_fp16)[name = tensor("linear_21_cast_fp16")]; + tensor input_117_cast_fp16 = add(x = linear_21_cast_fp16, y = hidden_states_39_cast_fp16)[name = tensor("input_117_cast_fp16")]; + tensor input_119_axes_0 = const()[name = tensor("input_119_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24381312)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24382656)))]; + tensor input_119_cast_fp16 = layer_norm(axes = input_119_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_1_norm2_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_1_norm2_weight_to_fp16, x = input_117_cast_fp16)[name = tensor("input_119_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24384000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24691264))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_22_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized, x = input_119_cast_fp16)[name = tensor("linear_22_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(24691456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25674560))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_23_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_23_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(25674752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26657856))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_24_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_24_cast_fp16")]; + tensor var_654 = const()[name = tensor("op_654"), val = tensor([1, -1, 10, 64])]; + tensor var_655_cast_fp16 = reshape(shape = var_654, x = linear_22_cast_fp16)[name = tensor("op_655_cast_fp16")]; + tensor var_657 = const()[name = tensor("op_657"), val = tensor([1, -1, 10, 64])]; + tensor var_658_cast_fp16 = reshape(shape = var_657, x = linear_23_cast_fp16)[name = tensor("op_658_cast_fp16")]; + tensor var_660 = const()[name = tensor("op_660"), val = tensor([1, -1, 10, 64])]; + tensor var_661_cast_fp16 = reshape(shape = var_660, x = linear_24_cast_fp16)[name = tensor("op_661_cast_fp16")]; + tensor value_15_perm_0 = const()[name = tensor("value_15_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_10_y_0_to_fp16 = const()[name = tensor("mul_10_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_10_cast_fp16 = mul(x = var_655_cast_fp16, y = mul_10_y_0_to_fp16)[name = tensor("mul_10_cast_fp16")]; + tensor matmul_3_transpose_y_0 = const()[name = tensor("matmul_3_transpose_y_0"), val = tensor(true)]; + tensor matmul_3_transpose_x_0 = const()[name = tensor("matmul_3_transpose_x_0"), val = tensor(false)]; + tensor transpose_278_perm_0 = const()[name = tensor("transpose_278_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_279_perm_0 = const()[name = tensor("transpose_279_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_675 = transpose(perm = transpose_279_perm_0, x = var_658_cast_fp16)[name = tensor("transpose_675")]; + tensor transpose_676 = transpose(perm = transpose_278_perm_0, x = mul_10_cast_fp16)[name = tensor("transpose_676")]; + tensor matmul_3_cast_fp16 = matmul(transpose_x = matmul_3_transpose_x_0, transpose_y = matmul_3_transpose_y_0, x = transpose_676, y = transpose_675)[name = tensor("matmul_3_cast_fp16")]; + tensor softmax_3_axis_0 = const()[name = tensor("softmax_3_axis_0"), val = tensor(-1)]; + tensor softmax_3_cast_fp16 = softmax(axis = softmax_3_axis_0, x = matmul_3_cast_fp16)[name = tensor("softmax_3_cast_fp16")]; + tensor hidden_states_49_transpose_x_0 = const()[name = tensor("hidden_states_49_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_49_transpose_y_0 = const()[name = tensor("hidden_states_49_transpose_y_0"), val = tensor(false)]; + tensor transpose_674 = transpose(perm = value_15_perm_0, x = var_661_cast_fp16)[name = tensor("transpose_674")]; + tensor hidden_states_49_cast_fp16 = matmul(transpose_x = hidden_states_49_transpose_x_0, transpose_y = hidden_states_49_transpose_y_0, x = softmax_3_cast_fp16, y = transpose_674)[name = tensor("hidden_states_49_cast_fp16")]; + tensor var_664_perm_0 = const()[name = tensor("op_664_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_668 = const()[name = tensor("op_668"), val = tensor([1, -1, 640])]; + tensor transpose_673 = transpose(perm = var_664_perm_0, x = hidden_states_49_cast_fp16)[name = tensor("transpose_673")]; + tensor hidden_states_51_cast_fp16 = reshape(shape = var_668, x = transpose_673)[name = tensor("hidden_states_51_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26658048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26965312))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26965504)))]; + tensor linear_25_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_51_cast_fp16)[name = tensor("linear_25_cast_fp16")]; + tensor input_125_cast_fp16 = add(x = linear_25_cast_fp16, y = input_117_cast_fp16)[name = tensor("input_125_cast_fp16")]; + tensor input_127_axes_0 = const()[name = tensor("input_127_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26966848)))]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26968192)))]; + tensor input_127_cast_fp16 = layer_norm(axes = input_127_axes_0, beta = down_blocks_1_attentions_0_transformer_blocks_1_norm3_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_0_transformer_blocks_1_norm3_weight_to_fp16, x = input_125_cast_fp16)[name = tensor("input_127_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(26969536))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29427200))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([5120, 640])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29427392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29431296))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([5120])]; + tensor linear_26_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_1_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized, x = input_127_cast_fp16)[name = tensor("linear_26_cast_fp16")]; + tensor var_690_split_sizes_0 = const()[name = tensor("op_690_split_sizes_0"), val = tensor([2560, 2560])]; + tensor var_690_axis_0 = const()[name = tensor("op_690_axis_0"), val = tensor(-1)]; + tensor var_690_cast_fp16_0, tensor var_690_cast_fp16_1 = split(axis = var_690_axis_0, split_sizes = var_690_split_sizes_0, x = linear_26_cast_fp16)[name = tensor("op_690_cast_fp16")]; + tensor var_692_mode_0 = const()[name = tensor("op_692_mode_0"), val = tensor("EXACT")]; + tensor var_692_cast_fp16 = gelu(mode = var_692_mode_0, x = var_690_cast_fp16_1)[name = tensor("op_692_cast_fp16")]; + tensor input_129_cast_fp16 = mul(x = var_690_cast_fp16_0, y = var_692_cast_fp16)[name = tensor("input_129_cast_fp16")]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(29431488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30660352))), name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized"), shape = tensor([640, 2560])]; + tensor down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30660544)))]; + tensor linear_27_cast_fp16 = linear(bias = down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16, weight = down_blocks_1_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized, x = input_129_cast_fp16)[name = tensor("linear_27_cast_fp16")]; + tensor input_133_cast_fp16 = add(x = linear_27_cast_fp16, y = input_125_cast_fp16)[name = tensor("input_133_cast_fp16")]; + tensor down_blocks_1_attentions_0_proj_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30661888))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30969152))), name = tensor("down_blocks_1_attentions_0_proj_out_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_0_proj_out_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_0_proj_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30969344)))]; + tensor linear_28_cast_fp16 = linear(bias = down_blocks_1_attentions_0_proj_out_bias_to_fp16, weight = down_blocks_1_attentions_0_proj_out_weight_to_fp16_palettized, x = input_133_cast_fp16)[name = tensor("linear_28_cast_fp16")]; + tensor var_702 = const()[name = tensor("op_702"), val = tensor([1, 64, 64, 640])]; + tensor var_703_cast_fp16 = reshape(shape = var_702, x = linear_28_cast_fp16)[name = tensor("op_703_cast_fp16")]; + tensor var_704 = const()[name = tensor("op_704"), val = tensor([0, 3, 1, 2])]; + tensor transpose_672 = transpose(perm = var_704, x = var_703_cast_fp16)[name = tensor("transpose_672")]; + tensor input_135_cast_fp16 = add(x = transpose_672, y = var_430_cast_fp16)[name = tensor("input_135_cast_fp16")]; + tensor reshape_28_shape_0 = const()[name = tensor("reshape_28_shape_0"), val = tensor([1, 32, 20, 64, 64])]; + tensor reshape_28_cast_fp16 = reshape(shape = reshape_28_shape_0, x = input_135_cast_fp16)[name = tensor("reshape_28_cast_fp16")]; + tensor reduce_mean_21_axes_0 = const()[name = tensor("reduce_mean_21_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_21_keep_dims_0 = const()[name = tensor("reduce_mean_21_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_21_cast_fp16 = reduce_mean(axes = reduce_mean_21_axes_0, keep_dims = reduce_mean_21_keep_dims_0, x = reshape_28_cast_fp16)[name = tensor("reduce_mean_21_cast_fp16")]; + tensor sub_14_cast_fp16 = sub(x = reshape_28_cast_fp16, y = reduce_mean_21_cast_fp16)[name = tensor("sub_14_cast_fp16")]; + tensor square_7_cast_fp16 = square(x = sub_14_cast_fp16)[name = tensor("square_7_cast_fp16")]; + tensor reduce_mean_23_axes_0 = const()[name = tensor("reduce_mean_23_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_23_keep_dims_0 = const()[name = tensor("reduce_mean_23_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_23_cast_fp16 = reduce_mean(axes = reduce_mean_23_axes_0, keep_dims = reduce_mean_23_keep_dims_0, x = square_7_cast_fp16)[name = tensor("reduce_mean_23_cast_fp16")]; + tensor add_14_y_0_to_fp16 = const()[name = tensor("add_14_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_14_cast_fp16 = add(x = reduce_mean_23_cast_fp16, y = add_14_y_0_to_fp16)[name = tensor("add_14_cast_fp16")]; + tensor sqrt_7_cast_fp16 = sqrt(x = add_14_cast_fp16)[name = tensor("sqrt_7_cast_fp16")]; + tensor real_div_7_cast_fp16 = real_div(x = sub_14_cast_fp16, y = sqrt_7_cast_fp16)[name = tensor("real_div_7_cast_fp16")]; + tensor reshape_29_shape_0 = const()[name = tensor("reshape_29_shape_0"), val = tensor([1, 640, 64, 64])]; + tensor reshape_29_cast_fp16 = reshape(shape = reshape_29_shape_0, x = real_div_7_cast_fp16)[name = tensor("reshape_29_cast_fp16")]; + tensor add_15_gamma_0_to_fp16 = const()[name = tensor("add_15_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30970688)))]; + tensor add_15_beta_0_to_fp16 = const()[name = tensor("add_15_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30972032)))]; + tensor add_15_epsilon_0_to_fp16 = const()[name = tensor("add_15_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_15_cast_fp16 = batch_norm(beta = add_15_beta_0_to_fp16, epsilon = add_15_epsilon_0_to_fp16, gamma = add_15_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_29_cast_fp16)[name = tensor("add_15_cast_fp16")]; + tensor input_139_cast_fp16 = silu(x = add_15_cast_fp16)[name = tensor("input_139_cast_fp16")]; + tensor var_719 = const()[name = tensor("op_719"), val = tensor([1, 1])]; + tensor var_721 = const()[name = tensor("op_721"), val = tensor([1, 1])]; + tensor hidden_states_63_pad_type_0 = const()[name = tensor("hidden_states_63_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_63_pad_0 = const()[name = tensor("hidden_states_63_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_1_resnets_1_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(30973376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33738240))), name = tensor("down_blocks_1_resnets_1_conv1_weight_to_fp16_palettized"), shape = tensor([640, 640, 3, 3])]; + tensor down_blocks_1_resnets_1_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33738432)))]; + tensor hidden_states_63_cast_fp16 = conv(bias = down_blocks_1_resnets_1_conv1_bias_to_fp16, dilations = var_721, groups = var_370, pad = hidden_states_63_pad_0, pad_type = hidden_states_63_pad_type_0, strides = var_719, weight = down_blocks_1_resnets_1_conv1_weight_to_fp16_palettized, x = input_139_cast_fp16)[name = tensor("hidden_states_63_cast_fp16")]; + tensor down_blocks_1_resnets_1_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(33739776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(34354240))), name = tensor("down_blocks_1_resnets_1_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([640, 1280])]; + tensor down_blocks_1_resnets_1_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_1_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(34354432)))]; + tensor linear_29_cast_fp16 = linear(bias = down_blocks_1_resnets_1_time_emb_proj_bias_to_fp16, weight = down_blocks_1_resnets_1_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_29_cast_fp16")]; + tensor var_730_axes_0 = const()[name = tensor("op_730_axes_0"), val = tensor([2])]; + tensor var_730_cast_fp16 = expand_dims(axes = var_730_axes_0, x = linear_29_cast_fp16)[name = tensor("op_730_cast_fp16")]; + tensor temb_7_axes_0 = const()[name = tensor("temb_7_axes_0"), val = tensor([3])]; + tensor temb_7_cast_fp16 = expand_dims(axes = temb_7_axes_0, x = var_730_cast_fp16)[name = tensor("temb_7_cast_fp16")]; + tensor input_143_cast_fp16 = add(x = hidden_states_63_cast_fp16, y = temb_7_cast_fp16)[name = tensor("input_143_cast_fp16")]; + tensor reshape_32_shape_0 = const()[name = tensor("reshape_32_shape_0"), val = tensor([1, 32, 20, 64, 64])]; + tensor reshape_32_cast_fp16 = reshape(shape = reshape_32_shape_0, x = input_143_cast_fp16)[name = tensor("reshape_32_cast_fp16")]; + tensor reduce_mean_24_axes_0 = const()[name = tensor("reduce_mean_24_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_24_keep_dims_0 = const()[name = tensor("reduce_mean_24_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_24_cast_fp16 = reduce_mean(axes = reduce_mean_24_axes_0, keep_dims = reduce_mean_24_keep_dims_0, x = reshape_32_cast_fp16)[name = tensor("reduce_mean_24_cast_fp16")]; + tensor sub_16_cast_fp16 = sub(x = reshape_32_cast_fp16, y = reduce_mean_24_cast_fp16)[name = tensor("sub_16_cast_fp16")]; + tensor square_8_cast_fp16 = square(x = sub_16_cast_fp16)[name = tensor("square_8_cast_fp16")]; + tensor reduce_mean_26_axes_0 = const()[name = tensor("reduce_mean_26_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_26_keep_dims_0 = const()[name = tensor("reduce_mean_26_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_26_cast_fp16 = reduce_mean(axes = reduce_mean_26_axes_0, keep_dims = reduce_mean_26_keep_dims_0, x = square_8_cast_fp16)[name = tensor("reduce_mean_26_cast_fp16")]; + tensor add_16_y_0_to_fp16 = const()[name = tensor("add_16_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_16_cast_fp16 = add(x = reduce_mean_26_cast_fp16, y = add_16_y_0_to_fp16)[name = tensor("add_16_cast_fp16")]; + tensor sqrt_8_cast_fp16 = sqrt(x = add_16_cast_fp16)[name = tensor("sqrt_8_cast_fp16")]; + tensor real_div_8_cast_fp16 = real_div(x = sub_16_cast_fp16, y = sqrt_8_cast_fp16)[name = tensor("real_div_8_cast_fp16")]; + tensor reshape_33_shape_0 = const()[name = tensor("reshape_33_shape_0"), val = tensor([1, 640, 64, 64])]; + tensor reshape_33_cast_fp16 = reshape(shape = reshape_33_shape_0, x = real_div_8_cast_fp16)[name = tensor("reshape_33_cast_fp16")]; + tensor add_17_gamma_0_to_fp16 = const()[name = tensor("add_17_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(34355776)))]; + tensor add_17_beta_0_to_fp16 = const()[name = tensor("add_17_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(34357120)))]; + tensor add_17_epsilon_0_to_fp16 = const()[name = tensor("add_17_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_17_cast_fp16 = batch_norm(beta = add_17_beta_0_to_fp16, epsilon = add_17_epsilon_0_to_fp16, gamma = add_17_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_33_cast_fp16)[name = tensor("add_17_cast_fp16")]; + tensor input_147_cast_fp16 = silu(x = add_17_cast_fp16)[name = tensor("input_147_cast_fp16")]; + tensor var_740 = const()[name = tensor("op_740"), val = tensor([1, 1])]; + tensor var_742 = const()[name = tensor("op_742"), val = tensor([1, 1])]; + tensor hidden_states_65_pad_type_0 = const()[name = tensor("hidden_states_65_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_65_pad_0 = const()[name = tensor("hidden_states_65_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_1_resnets_1_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(34358464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37123328))), name = tensor("down_blocks_1_resnets_1_conv2_weight_to_fp16_palettized"), shape = tensor([640, 640, 3, 3])]; + tensor down_blocks_1_resnets_1_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_1_resnets_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37123520)))]; + tensor hidden_states_65_cast_fp16 = conv(bias = down_blocks_1_resnets_1_conv2_bias_to_fp16, dilations = var_742, groups = var_370, pad = hidden_states_65_pad_0, pad_type = hidden_states_65_pad_type_0, strides = var_740, weight = down_blocks_1_resnets_1_conv2_weight_to_fp16_palettized, x = input_147_cast_fp16)[name = tensor("hidden_states_65_cast_fp16")]; + tensor var_745_cast_fp16 = add(x = input_135_cast_fp16, y = hidden_states_65_cast_fp16)[name = tensor("op_745_cast_fp16")]; + tensor reshape_36_shape_0 = const()[name = tensor("reshape_36_shape_0"), val = tensor([1, 32, 20, 64, 64])]; + tensor reshape_36_cast_fp16 = reshape(shape = reshape_36_shape_0, x = var_745_cast_fp16)[name = tensor("reshape_36_cast_fp16")]; + tensor reduce_mean_27_axes_0 = const()[name = tensor("reduce_mean_27_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_27_keep_dims_0 = const()[name = tensor("reduce_mean_27_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_27_cast_fp16 = reduce_mean(axes = reduce_mean_27_axes_0, keep_dims = reduce_mean_27_keep_dims_0, x = reshape_36_cast_fp16)[name = tensor("reduce_mean_27_cast_fp16")]; + tensor sub_18_cast_fp16 = sub(x = reshape_36_cast_fp16, y = reduce_mean_27_cast_fp16)[name = tensor("sub_18_cast_fp16")]; + tensor square_9_cast_fp16 = square(x = sub_18_cast_fp16)[name = tensor("square_9_cast_fp16")]; + tensor reduce_mean_29_axes_0 = const()[name = tensor("reduce_mean_29_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_29_keep_dims_0 = const()[name = tensor("reduce_mean_29_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_29_cast_fp16 = reduce_mean(axes = reduce_mean_29_axes_0, keep_dims = reduce_mean_29_keep_dims_0, x = square_9_cast_fp16)[name = tensor("reduce_mean_29_cast_fp16")]; + tensor add_18_y_0_to_fp16 = const()[name = tensor("add_18_y_0_to_fp16"), val = tensor(0x1.1p-20)]; + tensor add_18_cast_fp16 = add(x = reduce_mean_29_cast_fp16, y = add_18_y_0_to_fp16)[name = tensor("add_18_cast_fp16")]; + tensor sqrt_9_cast_fp16 = sqrt(x = add_18_cast_fp16)[name = tensor("sqrt_9_cast_fp16")]; + tensor real_div_9_cast_fp16 = real_div(x = sub_18_cast_fp16, y = sqrt_9_cast_fp16)[name = tensor("real_div_9_cast_fp16")]; + tensor reshape_37_shape_0 = const()[name = tensor("reshape_37_shape_0"), val = tensor([1, 640, 64, 64])]; + tensor reshape_37_cast_fp16 = reshape(shape = reshape_37_shape_0, x = real_div_9_cast_fp16)[name = tensor("reshape_37_cast_fp16")]; + tensor add_19_gamma_0_to_fp16 = const()[name = tensor("add_19_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37124864)))]; + tensor add_19_beta_0_to_fp16 = const()[name = tensor("add_19_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37126208)))]; + tensor add_19_epsilon_0_to_fp16 = const()[name = tensor("add_19_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_19_cast_fp16 = batch_norm(beta = add_19_beta_0_to_fp16, epsilon = add_19_epsilon_0_to_fp16, gamma = add_19_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_37_cast_fp16)[name = tensor("add_19_cast_fp16")]; + tensor var_767 = const()[name = tensor("op_767"), val = tensor([0, 2, 3, 1])]; + tensor var_771 = const()[name = tensor("op_771"), val = tensor([1, 4096, 640])]; + tensor transpose_671 = transpose(perm = var_767, x = add_19_cast_fp16)[name = tensor("transpose_671")]; + tensor input_151_cast_fp16 = reshape(shape = var_771, x = transpose_671)[name = tensor("input_151_cast_fp16")]; + tensor down_blocks_1_attentions_1_proj_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37127552))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37434816))), name = tensor("down_blocks_1_attentions_1_proj_in_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_proj_in_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_proj_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37435008)))]; + tensor linear_30_cast_fp16 = linear(bias = down_blocks_1_attentions_1_proj_in_bias_to_fp16, weight = down_blocks_1_attentions_1_proj_in_weight_to_fp16_palettized, x = input_151_cast_fp16)[name = tensor("linear_30_cast_fp16")]; + tensor hidden_states_73_axes_0 = const()[name = tensor("hidden_states_73_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37436352)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37437696)))]; + tensor hidden_states_73_cast_fp16 = layer_norm(axes = hidden_states_73_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_0_norm1_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_0_norm1_weight_to_fp16, x = linear_30_cast_fp16)[name = tensor("hidden_states_73_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37439040))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37746304))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_31_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_73_cast_fp16)[name = tensor("linear_31_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(37746496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38053760))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_32_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_73_cast_fp16)[name = tensor("linear_32_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38053952))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38361216))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_33_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_73_cast_fp16)[name = tensor("linear_33_cast_fp16")]; + tensor var_804 = const()[name = tensor("op_804"), val = tensor([1, -1, 10, 64])]; + tensor var_805_cast_fp16 = reshape(shape = var_804, x = linear_31_cast_fp16)[name = tensor("op_805_cast_fp16")]; + tensor var_807 = const()[name = tensor("op_807"), val = tensor([1, -1, 10, 64])]; + tensor var_808_cast_fp16 = reshape(shape = var_807, x = linear_32_cast_fp16)[name = tensor("op_808_cast_fp16")]; + tensor var_810 = const()[name = tensor("op_810"), val = tensor([1, -1, 10, 64])]; + tensor var_811_cast_fp16 = reshape(shape = var_810, x = linear_33_cast_fp16)[name = tensor("op_811_cast_fp16")]; + tensor value_19_perm_0 = const()[name = tensor("value_19_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_14_y_0_to_fp16 = const()[name = tensor("mul_14_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_14_cast_fp16 = mul(x = var_805_cast_fp16, y = mul_14_y_0_to_fp16)[name = tensor("mul_14_cast_fp16")]; + tensor matmul_4_transpose_y_0 = const()[name = tensor("matmul_4_transpose_y_0"), val = tensor(true)]; + tensor matmul_4_transpose_x_0 = const()[name = tensor("matmul_4_transpose_x_0"), val = tensor(false)]; + tensor transpose_280_perm_0 = const()[name = tensor("transpose_280_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_281_perm_0 = const()[name = tensor("transpose_281_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_669 = transpose(perm = transpose_281_perm_0, x = var_808_cast_fp16)[name = tensor("transpose_669")]; + tensor transpose_670 = transpose(perm = transpose_280_perm_0, x = mul_14_cast_fp16)[name = tensor("transpose_670")]; + tensor matmul_4_cast_fp16 = matmul(transpose_x = matmul_4_transpose_x_0, transpose_y = matmul_4_transpose_y_0, x = transpose_670, y = transpose_669)[name = tensor("matmul_4_cast_fp16")]; + tensor softmax_4_axis_0 = const()[name = tensor("softmax_4_axis_0"), val = tensor(-1)]; + tensor softmax_4_cast_fp16 = softmax(axis = softmax_4_axis_0, x = matmul_4_cast_fp16)[name = tensor("softmax_4_cast_fp16")]; + tensor hidden_states_75_transpose_x_0 = const()[name = tensor("hidden_states_75_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_75_transpose_y_0 = const()[name = tensor("hidden_states_75_transpose_y_0"), val = tensor(false)]; + tensor transpose_668 = transpose(perm = value_19_perm_0, x = var_811_cast_fp16)[name = tensor("transpose_668")]; + tensor hidden_states_75_cast_fp16 = matmul(transpose_x = hidden_states_75_transpose_x_0, transpose_y = hidden_states_75_transpose_y_0, x = softmax_4_cast_fp16, y = transpose_668)[name = tensor("hidden_states_75_cast_fp16")]; + tensor var_814_perm_0 = const()[name = tensor("op_814_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_818 = const()[name = tensor("op_818"), val = tensor([1, -1, 640])]; + tensor transpose_667 = transpose(perm = var_814_perm_0, x = hidden_states_75_cast_fp16)[name = tensor("transpose_667")]; + tensor hidden_states_77_cast_fp16 = reshape(shape = var_818, x = transpose_667)[name = tensor("hidden_states_77_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38361408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38668672))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38668864)))]; + tensor linear_34_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_77_cast_fp16)[name = tensor("linear_34_cast_fp16")]; + tensor input_157_cast_fp16 = add(x = linear_34_cast_fp16, y = linear_30_cast_fp16)[name = tensor("input_157_cast_fp16")]; + tensor input_159_axes_0 = const()[name = tensor("input_159_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38670208)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38671552)))]; + tensor input_159_cast_fp16 = layer_norm(axes = input_159_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_0_norm2_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_0_norm2_weight_to_fp16, x = input_157_cast_fp16)[name = tensor("input_159_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38672896))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38980160))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_35_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized, x = input_159_cast_fp16)[name = tensor("linear_35_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(38980352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39963456))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_36_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_36_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(39963648))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(40946752))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_37_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_37_cast_fp16")]; + tensor var_850 = const()[name = tensor("op_850"), val = tensor([1, -1, 10, 64])]; + tensor var_851_cast_fp16 = reshape(shape = var_850, x = linear_35_cast_fp16)[name = tensor("op_851_cast_fp16")]; + tensor var_853 = const()[name = tensor("op_853"), val = tensor([1, -1, 10, 64])]; + tensor var_854_cast_fp16 = reshape(shape = var_853, x = linear_36_cast_fp16)[name = tensor("op_854_cast_fp16")]; + tensor var_856 = const()[name = tensor("op_856"), val = tensor([1, -1, 10, 64])]; + tensor var_857_cast_fp16 = reshape(shape = var_856, x = linear_37_cast_fp16)[name = tensor("op_857_cast_fp16")]; + tensor value_23_perm_0 = const()[name = tensor("value_23_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_15_y_0_to_fp16 = const()[name = tensor("mul_15_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_15_cast_fp16 = mul(x = var_851_cast_fp16, y = mul_15_y_0_to_fp16)[name = tensor("mul_15_cast_fp16")]; + tensor matmul_5_transpose_y_0 = const()[name = tensor("matmul_5_transpose_y_0"), val = tensor(true)]; + tensor matmul_5_transpose_x_0 = const()[name = tensor("matmul_5_transpose_x_0"), val = tensor(false)]; + tensor transpose_282_perm_0 = const()[name = tensor("transpose_282_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_283_perm_0 = const()[name = tensor("transpose_283_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_665 = transpose(perm = transpose_283_perm_0, x = var_854_cast_fp16)[name = tensor("transpose_665")]; + tensor transpose_666 = transpose(perm = transpose_282_perm_0, x = mul_15_cast_fp16)[name = tensor("transpose_666")]; + tensor matmul_5_cast_fp16 = matmul(transpose_x = matmul_5_transpose_x_0, transpose_y = matmul_5_transpose_y_0, x = transpose_666, y = transpose_665)[name = tensor("matmul_5_cast_fp16")]; + tensor softmax_5_axis_0 = const()[name = tensor("softmax_5_axis_0"), val = tensor(-1)]; + tensor softmax_5_cast_fp16 = softmax(axis = softmax_5_axis_0, x = matmul_5_cast_fp16)[name = tensor("softmax_5_cast_fp16")]; + tensor hidden_states_81_transpose_x_0 = const()[name = tensor("hidden_states_81_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_81_transpose_y_0 = const()[name = tensor("hidden_states_81_transpose_y_0"), val = tensor(false)]; + tensor transpose_664 = transpose(perm = value_23_perm_0, x = var_857_cast_fp16)[name = tensor("transpose_664")]; + tensor hidden_states_81_cast_fp16 = matmul(transpose_x = hidden_states_81_transpose_x_0, transpose_y = hidden_states_81_transpose_y_0, x = softmax_5_cast_fp16, y = transpose_664)[name = tensor("hidden_states_81_cast_fp16")]; + tensor var_860_perm_0 = const()[name = tensor("op_860_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_864 = const()[name = tensor("op_864"), val = tensor([1, -1, 640])]; + tensor transpose_663 = transpose(perm = var_860_perm_0, x = hidden_states_81_cast_fp16)[name = tensor("transpose_663")]; + tensor hidden_states_83_cast_fp16 = reshape(shape = var_864, x = transpose_663)[name = tensor("hidden_states_83_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(40946944))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(41254208))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(41254400)))]; + tensor linear_38_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_83_cast_fp16)[name = tensor("linear_38_cast_fp16")]; + tensor input_165_cast_fp16 = add(x = linear_38_cast_fp16, y = input_157_cast_fp16)[name = tensor("input_165_cast_fp16")]; + tensor input_167_axes_0 = const()[name = tensor("input_167_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(41255744)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(41257088)))]; + tensor input_167_cast_fp16 = layer_norm(axes = input_167_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_0_norm3_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_0_norm3_weight_to_fp16, x = input_165_cast_fp16)[name = tensor("input_167_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(41258432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43716096))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([5120, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43716288))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43720192))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([5120])]; + tensor linear_39_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_1_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized, x = input_167_cast_fp16)[name = tensor("linear_39_cast_fp16")]; + tensor var_886_split_sizes_0 = const()[name = tensor("op_886_split_sizes_0"), val = tensor([2560, 2560])]; + tensor var_886_axis_0 = const()[name = tensor("op_886_axis_0"), val = tensor(-1)]; + tensor var_886_cast_fp16_0, tensor var_886_cast_fp16_1 = split(axis = var_886_axis_0, split_sizes = var_886_split_sizes_0, x = linear_39_cast_fp16)[name = tensor("op_886_cast_fp16")]; + tensor var_888_mode_0 = const()[name = tensor("op_888_mode_0"), val = tensor("EXACT")]; + tensor var_888_cast_fp16 = gelu(mode = var_888_mode_0, x = var_886_cast_fp16_1)[name = tensor("op_888_cast_fp16")]; + tensor input_169_cast_fp16 = mul(x = var_886_cast_fp16_0, y = var_888_cast_fp16)[name = tensor("input_169_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(43720384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44949248))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized"), shape = tensor([640, 2560])]; + tensor down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44949440)))]; + tensor linear_40_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized, x = input_169_cast_fp16)[name = tensor("linear_40_cast_fp16")]; + tensor hidden_states_91_cast_fp16 = add(x = linear_40_cast_fp16, y = input_165_cast_fp16)[name = tensor("hidden_states_91_cast_fp16")]; + tensor hidden_states_93_axes_0 = const()[name = tensor("hidden_states_93_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44950784)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44952128)))]; + tensor hidden_states_93_cast_fp16 = layer_norm(axes = hidden_states_93_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_1_norm1_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_1_norm1_weight_to_fp16, x = hidden_states_91_cast_fp16)[name = tensor("hidden_states_93_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(44953472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45260736))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_41_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_93_cast_fp16)[name = tensor("linear_41_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45260928))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45568192))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_42_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_93_cast_fp16)[name = tensor("linear_42_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45568384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45875648))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_43_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_93_cast_fp16)[name = tensor("linear_43_cast_fp16")]; + tensor var_923 = const()[name = tensor("op_923"), val = tensor([1, -1, 10, 64])]; + tensor var_924_cast_fp16 = reshape(shape = var_923, x = linear_41_cast_fp16)[name = tensor("op_924_cast_fp16")]; + tensor var_926 = const()[name = tensor("op_926"), val = tensor([1, -1, 10, 64])]; + tensor var_927_cast_fp16 = reshape(shape = var_926, x = linear_42_cast_fp16)[name = tensor("op_927_cast_fp16")]; + tensor var_929 = const()[name = tensor("op_929"), val = tensor([1, -1, 10, 64])]; + tensor var_930_cast_fp16 = reshape(shape = var_929, x = linear_43_cast_fp16)[name = tensor("op_930_cast_fp16")]; + tensor value_27_perm_0 = const()[name = tensor("value_27_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_16_y_0_to_fp16 = const()[name = tensor("mul_16_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_16_cast_fp16 = mul(x = var_924_cast_fp16, y = mul_16_y_0_to_fp16)[name = tensor("mul_16_cast_fp16")]; + tensor matmul_6_transpose_y_0 = const()[name = tensor("matmul_6_transpose_y_0"), val = tensor(true)]; + tensor matmul_6_transpose_x_0 = const()[name = tensor("matmul_6_transpose_x_0"), val = tensor(false)]; + tensor transpose_284_perm_0 = const()[name = tensor("transpose_284_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_285_perm_0 = const()[name = tensor("transpose_285_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_661 = transpose(perm = transpose_285_perm_0, x = var_927_cast_fp16)[name = tensor("transpose_661")]; + tensor transpose_662 = transpose(perm = transpose_284_perm_0, x = mul_16_cast_fp16)[name = tensor("transpose_662")]; + tensor matmul_6_cast_fp16 = matmul(transpose_x = matmul_6_transpose_x_0, transpose_y = matmul_6_transpose_y_0, x = transpose_662, y = transpose_661)[name = tensor("matmul_6_cast_fp16")]; + tensor softmax_6_axis_0 = const()[name = tensor("softmax_6_axis_0"), val = tensor(-1)]; + tensor softmax_6_cast_fp16 = softmax(axis = softmax_6_axis_0, x = matmul_6_cast_fp16)[name = tensor("softmax_6_cast_fp16")]; + tensor hidden_states_95_transpose_x_0 = const()[name = tensor("hidden_states_95_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_95_transpose_y_0 = const()[name = tensor("hidden_states_95_transpose_y_0"), val = tensor(false)]; + tensor transpose_660 = transpose(perm = value_27_perm_0, x = var_930_cast_fp16)[name = tensor("transpose_660")]; + tensor hidden_states_95_cast_fp16 = matmul(transpose_x = hidden_states_95_transpose_x_0, transpose_y = hidden_states_95_transpose_y_0, x = softmax_6_cast_fp16, y = transpose_660)[name = tensor("hidden_states_95_cast_fp16")]; + tensor var_933_perm_0 = const()[name = tensor("op_933_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_937 = const()[name = tensor("op_937"), val = tensor([1, -1, 640])]; + tensor transpose_659 = transpose(perm = var_933_perm_0, x = hidden_states_95_cast_fp16)[name = tensor("transpose_659")]; + tensor hidden_states_97_cast_fp16 = reshape(shape = var_937, x = transpose_659)[name = tensor("hidden_states_97_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(45875840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46183104))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46183296)))]; + tensor linear_44_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_97_cast_fp16)[name = tensor("linear_44_cast_fp16")]; + tensor input_177_cast_fp16 = add(x = linear_44_cast_fp16, y = hidden_states_91_cast_fp16)[name = tensor("input_177_cast_fp16")]; + tensor input_179_axes_0 = const()[name = tensor("input_179_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46184640)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46185984)))]; + tensor input_179_cast_fp16 = layer_norm(axes = input_179_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_1_norm2_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_1_norm2_weight_to_fp16, x = input_177_cast_fp16)[name = tensor("input_179_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46187328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46494592))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor linear_45_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized, x = input_179_cast_fp16)[name = tensor("linear_45_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(46494784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47477888))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_46_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_46_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(47478080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48461184))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([640, 2048])]; + tensor linear_47_cast_fp16 = linear(bias = add_11_mean_0_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_47_cast_fp16")]; + tensor var_969 = const()[name = tensor("op_969"), val = tensor([1, -1, 10, 64])]; + tensor var_970_cast_fp16 = reshape(shape = var_969, x = linear_45_cast_fp16)[name = tensor("op_970_cast_fp16")]; + tensor var_972 = const()[name = tensor("op_972"), val = tensor([1, -1, 10, 64])]; + tensor var_973_cast_fp16 = reshape(shape = var_972, x = linear_46_cast_fp16)[name = tensor("op_973_cast_fp16")]; + tensor var_975 = const()[name = tensor("op_975"), val = tensor([1, -1, 10, 64])]; + tensor var_976_cast_fp16 = reshape(shape = var_975, x = linear_47_cast_fp16)[name = tensor("op_976_cast_fp16")]; + tensor value_31_perm_0 = const()[name = tensor("value_31_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_17_y_0_to_fp16 = const()[name = tensor("mul_17_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_17_cast_fp16 = mul(x = var_970_cast_fp16, y = mul_17_y_0_to_fp16)[name = tensor("mul_17_cast_fp16")]; + tensor matmul_7_transpose_y_0 = const()[name = tensor("matmul_7_transpose_y_0"), val = tensor(true)]; + tensor matmul_7_transpose_x_0 = const()[name = tensor("matmul_7_transpose_x_0"), val = tensor(false)]; + tensor transpose_286_perm_0 = const()[name = tensor("transpose_286_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_287_perm_0 = const()[name = tensor("transpose_287_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_657 = transpose(perm = transpose_287_perm_0, x = var_973_cast_fp16)[name = tensor("transpose_657")]; + tensor transpose_658 = transpose(perm = transpose_286_perm_0, x = mul_17_cast_fp16)[name = tensor("transpose_658")]; + tensor matmul_7_cast_fp16 = matmul(transpose_x = matmul_7_transpose_x_0, transpose_y = matmul_7_transpose_y_0, x = transpose_658, y = transpose_657)[name = tensor("matmul_7_cast_fp16")]; + tensor softmax_7_axis_0 = const()[name = tensor("softmax_7_axis_0"), val = tensor(-1)]; + tensor softmax_7_cast_fp16 = softmax(axis = softmax_7_axis_0, x = matmul_7_cast_fp16)[name = tensor("softmax_7_cast_fp16")]; + tensor hidden_states_101_transpose_x_0 = const()[name = tensor("hidden_states_101_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_101_transpose_y_0 = const()[name = tensor("hidden_states_101_transpose_y_0"), val = tensor(false)]; + tensor transpose_656 = transpose(perm = value_31_perm_0, x = var_976_cast_fp16)[name = tensor("transpose_656")]; + tensor hidden_states_101_cast_fp16 = matmul(transpose_x = hidden_states_101_transpose_x_0, transpose_y = hidden_states_101_transpose_y_0, x = softmax_7_cast_fp16, y = transpose_656)[name = tensor("hidden_states_101_cast_fp16")]; + tensor var_979_perm_0 = const()[name = tensor("op_979_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_983 = const()[name = tensor("op_983"), val = tensor([1, -1, 640])]; + tensor transpose_655 = transpose(perm = var_979_perm_0, x = hidden_states_101_cast_fp16)[name = tensor("transpose_655")]; + tensor hidden_states_103_cast_fp16 = reshape(shape = var_983, x = transpose_655)[name = tensor("hidden_states_103_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48461376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48768640))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48768832)))]; + tensor linear_48_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_103_cast_fp16)[name = tensor("linear_48_cast_fp16")]; + tensor input_185_cast_fp16 = add(x = linear_48_cast_fp16, y = input_177_cast_fp16)[name = tensor("input_185_cast_fp16")]; + tensor input_187_axes_0 = const()[name = tensor("input_187_axes_0"), val = tensor([-1])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48770176)))]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48771520)))]; + tensor input_187_cast_fp16 = layer_norm(axes = input_187_axes_0, beta = down_blocks_1_attentions_1_transformer_blocks_1_norm3_bias_to_fp16, epsilon = var_366_to_fp16, gamma = down_blocks_1_attentions_1_transformer_blocks_1_norm3_weight_to_fp16, x = input_185_cast_fp16)[name = tensor("input_187_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(48772864))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51230528))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([5120, 640])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51230720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51234624))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([5120])]; + tensor linear_49_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_1_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized, x = input_187_cast_fp16)[name = tensor("linear_49_cast_fp16")]; + tensor var_1005_split_sizes_0 = const()[name = tensor("op_1005_split_sizes_0"), val = tensor([2560, 2560])]; + tensor var_1005_axis_0 = const()[name = tensor("op_1005_axis_0"), val = tensor(-1)]; + tensor var_1005_cast_fp16_0, tensor var_1005_cast_fp16_1 = split(axis = var_1005_axis_0, split_sizes = var_1005_split_sizes_0, x = linear_49_cast_fp16)[name = tensor("op_1005_cast_fp16")]; + tensor var_1007_mode_0 = const()[name = tensor("op_1007_mode_0"), val = tensor("EXACT")]; + tensor var_1007_cast_fp16 = gelu(mode = var_1007_mode_0, x = var_1005_cast_fp16_1)[name = tensor("op_1007_cast_fp16")]; + tensor input_189_cast_fp16 = mul(x = var_1005_cast_fp16_0, y = var_1007_cast_fp16)[name = tensor("input_189_cast_fp16")]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(51234816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52463680))), name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized"), shape = tensor([640, 2560])]; + tensor down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52463872)))]; + tensor linear_50_cast_fp16 = linear(bias = down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16, weight = down_blocks_1_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized, x = input_189_cast_fp16)[name = tensor("linear_50_cast_fp16")]; + tensor input_193_cast_fp16 = add(x = linear_50_cast_fp16, y = input_185_cast_fp16)[name = tensor("input_193_cast_fp16")]; + tensor down_blocks_1_attentions_1_proj_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52465216))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52772480))), name = tensor("down_blocks_1_attentions_1_proj_out_weight_to_fp16_palettized"), shape = tensor([640, 640])]; + tensor down_blocks_1_attentions_1_proj_out_bias_to_fp16 = const()[name = tensor("down_blocks_1_attentions_1_proj_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52772672)))]; + tensor linear_51_cast_fp16 = linear(bias = down_blocks_1_attentions_1_proj_out_bias_to_fp16, weight = down_blocks_1_attentions_1_proj_out_weight_to_fp16_palettized, x = input_193_cast_fp16)[name = tensor("linear_51_cast_fp16")]; + tensor var_1017 = const()[name = tensor("op_1017"), val = tensor([1, 64, 64, 640])]; + tensor var_1018_cast_fp16 = reshape(shape = var_1017, x = linear_51_cast_fp16)[name = tensor("op_1018_cast_fp16")]; + tensor var_1019 = const()[name = tensor("op_1019"), val = tensor([0, 3, 1, 2])]; + tensor transpose_654 = transpose(perm = var_1019, x = var_1018_cast_fp16)[name = tensor("transpose_654")]; + tensor hidden_states_115_cast_fp16 = add(x = transpose_654, y = var_745_cast_fp16)[name = tensor("hidden_states_115_cast_fp16")]; + tensor var_1026 = const()[name = tensor("op_1026"), val = tensor([2, 2])]; + tensor var_1028 = const()[name = tensor("op_1028"), val = tensor([1, 1])]; + tensor input_195_pad_type_0 = const()[name = tensor("input_195_pad_type_0"), val = tensor("custom")]; + tensor input_195_pad_0 = const()[name = tensor("input_195_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_1_downsamplers_0_conv_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(52774016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(55538880))), name = tensor("down_blocks_1_downsamplers_0_conv_weight_to_fp16_palettized"), shape = tensor([640, 640, 3, 3])]; + tensor down_blocks_1_downsamplers_0_conv_bias_to_fp16 = const()[name = tensor("down_blocks_1_downsamplers_0_conv_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(55539072)))]; + tensor input_195_cast_fp16 = conv(bias = down_blocks_1_downsamplers_0_conv_bias_to_fp16, dilations = var_1028, groups = var_370, pad = input_195_pad_0, pad_type = input_195_pad_type_0, strides = var_1026, weight = down_blocks_1_downsamplers_0_conv_weight_to_fp16_palettized, x = hidden_states_115_cast_fp16)[name = tensor("input_195_cast_fp16")]; + tensor var_1048 = const()[name = tensor("op_1048"), val = tensor(1)]; + tensor reshape_40_shape_0 = const()[name = tensor("reshape_40_shape_0"), val = tensor([1, 32, 20, 32, 32])]; + tensor reshape_40_cast_fp16 = reshape(shape = reshape_40_shape_0, x = input_195_cast_fp16)[name = tensor("reshape_40_cast_fp16")]; + tensor reduce_mean_30_axes_0 = const()[name = tensor("reduce_mean_30_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_30_keep_dims_0 = const()[name = tensor("reduce_mean_30_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_30_cast_fp16 = reduce_mean(axes = reduce_mean_30_axes_0, keep_dims = reduce_mean_30_keep_dims_0, x = reshape_40_cast_fp16)[name = tensor("reduce_mean_30_cast_fp16")]; + tensor sub_20_cast_fp16 = sub(x = reshape_40_cast_fp16, y = reduce_mean_30_cast_fp16)[name = tensor("sub_20_cast_fp16")]; + tensor square_10_cast_fp16 = square(x = sub_20_cast_fp16)[name = tensor("square_10_cast_fp16")]; + tensor reduce_mean_32_axes_0 = const()[name = tensor("reduce_mean_32_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_32_keep_dims_0 = const()[name = tensor("reduce_mean_32_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_32_cast_fp16 = reduce_mean(axes = reduce_mean_32_axes_0, keep_dims = reduce_mean_32_keep_dims_0, x = square_10_cast_fp16)[name = tensor("reduce_mean_32_cast_fp16")]; + tensor add_20_y_0_to_fp16 = const()[name = tensor("add_20_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_20_cast_fp16 = add(x = reduce_mean_32_cast_fp16, y = add_20_y_0_to_fp16)[name = tensor("add_20_cast_fp16")]; + tensor sqrt_10_cast_fp16 = sqrt(x = add_20_cast_fp16)[name = tensor("sqrt_10_cast_fp16")]; + tensor real_div_10_cast_fp16 = real_div(x = sub_20_cast_fp16, y = sqrt_10_cast_fp16)[name = tensor("real_div_10_cast_fp16")]; + tensor reshape_41_shape_0 = const()[name = tensor("reshape_41_shape_0"), val = tensor([1, 640, 32, 32])]; + tensor reshape_41_cast_fp16 = reshape(shape = reshape_41_shape_0, x = real_div_10_cast_fp16)[name = tensor("reshape_41_cast_fp16")]; + tensor add_21_gamma_0_to_fp16 = const()[name = tensor("add_21_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(55540416)))]; + tensor add_21_beta_0_to_fp16 = const()[name = tensor("add_21_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(55541760)))]; + tensor add_21_epsilon_0_to_fp16 = const()[name = tensor("add_21_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_21_cast_fp16 = batch_norm(beta = add_21_beta_0_to_fp16, epsilon = add_21_epsilon_0_to_fp16, gamma = add_21_gamma_0_to_fp16, mean = add_11_mean_0_to_fp16, variance = add_11_variance_0_to_fp16, x = reshape_41_cast_fp16)[name = tensor("add_21_cast_fp16")]; + tensor input_199_cast_fp16 = silu(x = add_21_cast_fp16)[name = tensor("input_199_cast_fp16")]; + tensor var_1073 = const()[name = tensor("op_1073"), val = tensor([1, 1])]; + tensor var_1075 = const()[name = tensor("op_1075"), val = tensor([1, 1])]; + tensor hidden_states_117_pad_type_0 = const()[name = tensor("hidden_states_117_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_117_pad_0 = const()[name = tensor("hidden_states_117_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_2_resnets_0_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(55543104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(61072768))), name = tensor("down_blocks_2_resnets_0_conv1_weight_to_fp16_palettized"), shape = tensor([1280, 640, 3, 3])]; + tensor down_blocks_2_resnets_0_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(61072960)))]; + tensor hidden_states_117_cast_fp16 = conv(bias = down_blocks_2_resnets_0_conv1_bias_to_fp16, dilations = var_1075, groups = var_1048, pad = hidden_states_117_pad_0, pad_type = hidden_states_117_pad_type_0, strides = var_1073, weight = down_blocks_2_resnets_0_conv1_weight_to_fp16_palettized, x = input_199_cast_fp16)[name = tensor("hidden_states_117_cast_fp16")]; + tensor down_blocks_2_resnets_0_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(61075584))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62304448))), name = tensor("down_blocks_2_resnets_0_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_resnets_0_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_0_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62304640)))]; + tensor linear_52_cast_fp16 = linear(bias = down_blocks_2_resnets_0_time_emb_proj_bias_to_fp16, weight = down_blocks_2_resnets_0_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_52_cast_fp16")]; + tensor var_1084_axes_0 = const()[name = tensor("op_1084_axes_0"), val = tensor([2])]; + tensor var_1084_cast_fp16 = expand_dims(axes = var_1084_axes_0, x = linear_52_cast_fp16)[name = tensor("op_1084_cast_fp16")]; + tensor temb_9_axes_0 = const()[name = tensor("temb_9_axes_0"), val = tensor([3])]; + tensor temb_9_cast_fp16 = expand_dims(axes = temb_9_axes_0, x = var_1084_cast_fp16)[name = tensor("temb_9_cast_fp16")]; + tensor input_203_cast_fp16 = add(x = hidden_states_117_cast_fp16, y = temb_9_cast_fp16)[name = tensor("input_203_cast_fp16")]; + tensor reshape_44_shape_0 = const()[name = tensor("reshape_44_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_44_cast_fp16 = reshape(shape = reshape_44_shape_0, x = input_203_cast_fp16)[name = tensor("reshape_44_cast_fp16")]; + tensor reduce_mean_33_axes_0 = const()[name = tensor("reduce_mean_33_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_33_keep_dims_0 = const()[name = tensor("reduce_mean_33_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_33_cast_fp16 = reduce_mean(axes = reduce_mean_33_axes_0, keep_dims = reduce_mean_33_keep_dims_0, x = reshape_44_cast_fp16)[name = tensor("reduce_mean_33_cast_fp16")]; + tensor sub_22_cast_fp16 = sub(x = reshape_44_cast_fp16, y = reduce_mean_33_cast_fp16)[name = tensor("sub_22_cast_fp16")]; + tensor square_11_cast_fp16 = square(x = sub_22_cast_fp16)[name = tensor("square_11_cast_fp16")]; + tensor reduce_mean_35_axes_0 = const()[name = tensor("reduce_mean_35_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_35_keep_dims_0 = const()[name = tensor("reduce_mean_35_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_35_cast_fp16 = reduce_mean(axes = reduce_mean_35_axes_0, keep_dims = reduce_mean_35_keep_dims_0, x = square_11_cast_fp16)[name = tensor("reduce_mean_35_cast_fp16")]; + tensor add_22_y_0_to_fp16 = const()[name = tensor("add_22_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_22_cast_fp16 = add(x = reduce_mean_35_cast_fp16, y = add_22_y_0_to_fp16)[name = tensor("add_22_cast_fp16")]; + tensor sqrt_11_cast_fp16 = sqrt(x = add_22_cast_fp16)[name = tensor("sqrt_11_cast_fp16")]; + tensor real_div_11_cast_fp16 = real_div(x = sub_22_cast_fp16, y = sqrt_11_cast_fp16)[name = tensor("real_div_11_cast_fp16")]; + tensor reshape_45_shape_0 = const()[name = tensor("reshape_45_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_45_cast_fp16 = reshape(shape = reshape_45_shape_0, x = real_div_11_cast_fp16)[name = tensor("reshape_45_cast_fp16")]; + tensor add_23_mean_0_to_fp16 = const()[name = tensor("add_23_mean_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62307264)))]; + tensor add_23_variance_0_to_fp16 = const()[name = tensor("add_23_variance_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62309888)))]; + tensor add_23_gamma_0_to_fp16 = const()[name = tensor("add_23_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62312512)))]; + tensor add_23_beta_0_to_fp16 = const()[name = tensor("add_23_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62315136)))]; + tensor add_23_epsilon_0_to_fp16 = const()[name = tensor("add_23_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_23_cast_fp16 = batch_norm(beta = add_23_beta_0_to_fp16, epsilon = add_23_epsilon_0_to_fp16, gamma = add_23_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_45_cast_fp16)[name = tensor("add_23_cast_fp16")]; + tensor input_207_cast_fp16 = silu(x = add_23_cast_fp16)[name = tensor("input_207_cast_fp16")]; + tensor var_1094 = const()[name = tensor("op_1094"), val = tensor([1, 1])]; + tensor var_1096 = const()[name = tensor("op_1096"), val = tensor([1, 1])]; + tensor hidden_states_119_pad_type_0 = const()[name = tensor("hidden_states_119_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_119_pad_0 = const()[name = tensor("hidden_states_119_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_2_resnets_0_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(62317760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73377024))), name = tensor("down_blocks_2_resnets_0_conv2_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor down_blocks_2_resnets_0_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73377216)))]; + tensor hidden_states_119_cast_fp16 = conv(bias = down_blocks_2_resnets_0_conv2_bias_to_fp16, dilations = var_1096, groups = var_1048, pad = hidden_states_119_pad_0, pad_type = hidden_states_119_pad_type_0, strides = var_1094, weight = down_blocks_2_resnets_0_conv2_weight_to_fp16_palettized, x = input_207_cast_fp16)[name = tensor("hidden_states_119_cast_fp16")]; + tensor var_1101 = const()[name = tensor("op_1101"), val = tensor([1, 1])]; + tensor var_1103 = const()[name = tensor("op_1103"), val = tensor([1, 1])]; + tensor input_tensor_pad_type_0 = const()[name = tensor("input_tensor_pad_type_0"), val = tensor("custom")]; + tensor input_tensor_pad_0 = const()[name = tensor("input_tensor_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor down_blocks_2_resnets_0_conv_shortcut_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73379840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73994304))), name = tensor("down_blocks_2_resnets_0_conv_shortcut_weight_to_fp16_palettized"), shape = tensor([1280, 640, 1, 1])]; + tensor down_blocks_2_resnets_0_conv_shortcut_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_0_conv_shortcut_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73994496)))]; + tensor input_tensor_cast_fp16 = conv(bias = down_blocks_2_resnets_0_conv_shortcut_bias_to_fp16, dilations = var_1103, groups = var_1048, pad = input_tensor_pad_0, pad_type = input_tensor_pad_type_0, strides = var_1101, weight = down_blocks_2_resnets_0_conv_shortcut_weight_to_fp16_palettized, x = input_195_cast_fp16)[name = tensor("input_tensor_cast_fp16")]; + tensor var_1106_cast_fp16 = add(x = input_tensor_cast_fp16, y = hidden_states_119_cast_fp16)[name = tensor("op_1106_cast_fp16")]; + tensor reshape_48_shape_0 = const()[name = tensor("reshape_48_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_48_cast_fp16 = reshape(shape = reshape_48_shape_0, x = var_1106_cast_fp16)[name = tensor("reshape_48_cast_fp16")]; + tensor reduce_mean_36_axes_0 = const()[name = tensor("reduce_mean_36_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_36_keep_dims_0 = const()[name = tensor("reduce_mean_36_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_36_cast_fp16 = reduce_mean(axes = reduce_mean_36_axes_0, keep_dims = reduce_mean_36_keep_dims_0, x = reshape_48_cast_fp16)[name = tensor("reduce_mean_36_cast_fp16")]; + tensor sub_24_cast_fp16 = sub(x = reshape_48_cast_fp16, y = reduce_mean_36_cast_fp16)[name = tensor("sub_24_cast_fp16")]; + tensor square_12_cast_fp16 = square(x = sub_24_cast_fp16)[name = tensor("square_12_cast_fp16")]; + tensor reduce_mean_38_axes_0 = const()[name = tensor("reduce_mean_38_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_38_keep_dims_0 = const()[name = tensor("reduce_mean_38_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_38_cast_fp16 = reduce_mean(axes = reduce_mean_38_axes_0, keep_dims = reduce_mean_38_keep_dims_0, x = square_12_cast_fp16)[name = tensor("reduce_mean_38_cast_fp16")]; + tensor add_24_y_0_to_fp16 = const()[name = tensor("add_24_y_0_to_fp16"), val = tensor(0x1.1p-20)]; + tensor add_24_cast_fp16 = add(x = reduce_mean_38_cast_fp16, y = add_24_y_0_to_fp16)[name = tensor("add_24_cast_fp16")]; + tensor sqrt_12_cast_fp16 = sqrt(x = add_24_cast_fp16)[name = tensor("sqrt_12_cast_fp16")]; + tensor real_div_12_cast_fp16 = real_div(x = sub_24_cast_fp16, y = sqrt_12_cast_fp16)[name = tensor("real_div_12_cast_fp16")]; + tensor reshape_49_shape_0 = const()[name = tensor("reshape_49_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_49_cast_fp16 = reshape(shape = reshape_49_shape_0, x = real_div_12_cast_fp16)[name = tensor("reshape_49_cast_fp16")]; + tensor add_25_gamma_0_to_fp16 = const()[name = tensor("add_25_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73997120)))]; + tensor add_25_beta_0_to_fp16 = const()[name = tensor("add_25_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(73999744)))]; + tensor add_25_epsilon_0_to_fp16 = const()[name = tensor("add_25_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_25_cast_fp16 = batch_norm(beta = add_25_beta_0_to_fp16, epsilon = add_25_epsilon_0_to_fp16, gamma = add_25_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_49_cast_fp16)[name = tensor("add_25_cast_fp16")]; + tensor var_1144 = const()[name = tensor("op_1144"), val = tensor([0, 2, 3, 1])]; + tensor var_1148 = const()[name = tensor("op_1148"), val = tensor([1, 1024, 1280])]; + tensor transpose_653 = transpose(perm = var_1144, x = add_25_cast_fp16)[name = tensor("transpose_653")]; + tensor input_211_cast_fp16 = reshape(shape = var_1148, x = transpose_653)[name = tensor("input_211_cast_fp16")]; + tensor down_blocks_2_attentions_0_proj_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(74002368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(75231232))), name = tensor("down_blocks_2_attentions_0_proj_in_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_proj_in_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_proj_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(75231424)))]; + tensor linear_53_cast_fp16 = linear(bias = down_blocks_2_attentions_0_proj_in_bias_to_fp16, weight = down_blocks_2_attentions_0_proj_in_weight_to_fp16_palettized, x = input_211_cast_fp16)[name = tensor("linear_53_cast_fp16")]; + tensor hidden_states_127_axes_0 = const()[name = tensor("hidden_states_127_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(75234048)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(75236672)))]; + tensor var_1044_to_fp16 = const()[name = tensor("op_1044_to_fp16"), val = tensor(0x1.5p-17)]; + tensor hidden_states_127_cast_fp16 = layer_norm(axes = hidden_states_127_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_0_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_0_norm1_weight_to_fp16, x = linear_53_cast_fp16)[name = tensor("hidden_states_127_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(75239296))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(76468160))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_54_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_127_cast_fp16)[name = tensor("linear_54_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(76468352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(77697216))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_55_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_127_cast_fp16)[name = tensor("linear_55_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(77697408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(78926272))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_56_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_127_cast_fp16)[name = tensor("linear_56_cast_fp16")]; + tensor var_1181 = const()[name = tensor("op_1181"), val = tensor([1, -1, 20, 64])]; + tensor var_1182_cast_fp16 = reshape(shape = var_1181, x = linear_54_cast_fp16)[name = tensor("op_1182_cast_fp16")]; + tensor var_1184 = const()[name = tensor("op_1184"), val = tensor([1, -1, 20, 64])]; + tensor var_1185_cast_fp16 = reshape(shape = var_1184, x = linear_55_cast_fp16)[name = tensor("op_1185_cast_fp16")]; + tensor var_1187 = const()[name = tensor("op_1187"), val = tensor([1, -1, 20, 64])]; + tensor var_1188_cast_fp16 = reshape(shape = var_1187, x = linear_56_cast_fp16)[name = tensor("op_1188_cast_fp16")]; + tensor value_35_perm_0 = const()[name = tensor("value_35_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_21_y_0_to_fp16 = const()[name = tensor("mul_21_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_21_cast_fp16 = mul(x = var_1182_cast_fp16, y = mul_21_y_0_to_fp16)[name = tensor("mul_21_cast_fp16")]; + tensor matmul_8_transpose_y_0 = const()[name = tensor("matmul_8_transpose_y_0"), val = tensor(true)]; + tensor matmul_8_transpose_x_0 = const()[name = tensor("matmul_8_transpose_x_0"), val = tensor(false)]; + tensor transpose_288_perm_0 = const()[name = tensor("transpose_288_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_289_perm_0 = const()[name = tensor("transpose_289_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_651 = transpose(perm = transpose_289_perm_0, x = var_1185_cast_fp16)[name = tensor("transpose_651")]; + tensor transpose_652 = transpose(perm = transpose_288_perm_0, x = mul_21_cast_fp16)[name = tensor("transpose_652")]; + tensor matmul_8_cast_fp16 = matmul(transpose_x = matmul_8_transpose_x_0, transpose_y = matmul_8_transpose_y_0, x = transpose_652, y = transpose_651)[name = tensor("matmul_8_cast_fp16")]; + tensor softmax_8_axis_0 = const()[name = tensor("softmax_8_axis_0"), val = tensor(-1)]; + tensor softmax_8_cast_fp16 = softmax(axis = softmax_8_axis_0, x = matmul_8_cast_fp16)[name = tensor("softmax_8_cast_fp16")]; + tensor hidden_states_129_transpose_x_0 = const()[name = tensor("hidden_states_129_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_129_transpose_y_0 = const()[name = tensor("hidden_states_129_transpose_y_0"), val = tensor(false)]; + tensor transpose_650 = transpose(perm = value_35_perm_0, x = var_1188_cast_fp16)[name = tensor("transpose_650")]; + tensor hidden_states_129_cast_fp16 = matmul(transpose_x = hidden_states_129_transpose_x_0, transpose_y = hidden_states_129_transpose_y_0, x = softmax_8_cast_fp16, y = transpose_650)[name = tensor("hidden_states_129_cast_fp16")]; + tensor var_1191_perm_0 = const()[name = tensor("op_1191_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1195 = const()[name = tensor("op_1195"), val = tensor([1, -1, 1280])]; + tensor transpose_649 = transpose(perm = var_1191_perm_0, x = hidden_states_129_cast_fp16)[name = tensor("transpose_649")]; + tensor hidden_states_131_cast_fp16 = reshape(shape = var_1195, x = transpose_649)[name = tensor("hidden_states_131_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(78926464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(80155328))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(80155520)))]; + tensor linear_57_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_131_cast_fp16)[name = tensor("linear_57_cast_fp16")]; + tensor input_217_cast_fp16 = add(x = linear_57_cast_fp16, y = linear_53_cast_fp16)[name = tensor("input_217_cast_fp16")]; + tensor input_219_axes_0 = const()[name = tensor("input_219_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(80158144)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(80160768)))]; + tensor input_219_cast_fp16 = layer_norm(axes = input_219_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_0_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_0_norm2_weight_to_fp16, x = input_217_cast_fp16)[name = tensor("input_219_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(80163392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(81392256))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_58_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized, x = input_219_cast_fp16)[name = tensor("linear_58_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(81392448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(83358592))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_59_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_59_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(83358784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(85324928))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_60_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_60_cast_fp16")]; + tensor var_1227 = const()[name = tensor("op_1227"), val = tensor([1, -1, 20, 64])]; + tensor var_1228_cast_fp16 = reshape(shape = var_1227, x = linear_58_cast_fp16)[name = tensor("op_1228_cast_fp16")]; + tensor var_1230 = const()[name = tensor("op_1230"), val = tensor([1, -1, 20, 64])]; + tensor var_1231_cast_fp16 = reshape(shape = var_1230, x = linear_59_cast_fp16)[name = tensor("op_1231_cast_fp16")]; + tensor var_1233 = const()[name = tensor("op_1233"), val = tensor([1, -1, 20, 64])]; + tensor var_1234_cast_fp16 = reshape(shape = var_1233, x = linear_60_cast_fp16)[name = tensor("op_1234_cast_fp16")]; + tensor value_39_perm_0 = const()[name = tensor("value_39_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_22_y_0_to_fp16 = const()[name = tensor("mul_22_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_22_cast_fp16 = mul(x = var_1228_cast_fp16, y = mul_22_y_0_to_fp16)[name = tensor("mul_22_cast_fp16")]; + tensor matmul_9_transpose_y_0 = const()[name = tensor("matmul_9_transpose_y_0"), val = tensor(true)]; + tensor matmul_9_transpose_x_0 = const()[name = tensor("matmul_9_transpose_x_0"), val = tensor(false)]; + tensor transpose_290_perm_0 = const()[name = tensor("transpose_290_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_291_perm_0 = const()[name = tensor("transpose_291_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_647 = transpose(perm = transpose_291_perm_0, x = var_1231_cast_fp16)[name = tensor("transpose_647")]; + tensor transpose_648 = transpose(perm = transpose_290_perm_0, x = mul_22_cast_fp16)[name = tensor("transpose_648")]; + tensor matmul_9_cast_fp16 = matmul(transpose_x = matmul_9_transpose_x_0, transpose_y = matmul_9_transpose_y_0, x = transpose_648, y = transpose_647)[name = tensor("matmul_9_cast_fp16")]; + tensor softmax_9_axis_0 = const()[name = tensor("softmax_9_axis_0"), val = tensor(-1)]; + tensor softmax_9_cast_fp16 = softmax(axis = softmax_9_axis_0, x = matmul_9_cast_fp16)[name = tensor("softmax_9_cast_fp16")]; + tensor hidden_states_135_transpose_x_0 = const()[name = tensor("hidden_states_135_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_135_transpose_y_0 = const()[name = tensor("hidden_states_135_transpose_y_0"), val = tensor(false)]; + tensor transpose_646 = transpose(perm = value_39_perm_0, x = var_1234_cast_fp16)[name = tensor("transpose_646")]; + tensor hidden_states_135_cast_fp16 = matmul(transpose_x = hidden_states_135_transpose_x_0, transpose_y = hidden_states_135_transpose_y_0, x = softmax_9_cast_fp16, y = transpose_646)[name = tensor("hidden_states_135_cast_fp16")]; + tensor var_1237_perm_0 = const()[name = tensor("op_1237_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1241 = const()[name = tensor("op_1241"), val = tensor([1, -1, 1280])]; + tensor transpose_645 = transpose(perm = var_1237_perm_0, x = hidden_states_135_cast_fp16)[name = tensor("transpose_645")]; + tensor hidden_states_137_cast_fp16 = reshape(shape = var_1241, x = transpose_645)[name = tensor("hidden_states_137_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(85325120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(86553984))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(86554176)))]; + tensor linear_61_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_137_cast_fp16)[name = tensor("linear_61_cast_fp16")]; + tensor input_225_cast_fp16 = add(x = linear_61_cast_fp16, y = input_217_cast_fp16)[name = tensor("input_225_cast_fp16")]; + tensor input_227_axes_0 = const()[name = tensor("input_227_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(86556800)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(86559424)))]; + tensor input_227_cast_fp16 = layer_norm(axes = input_227_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_0_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_0_norm3_weight_to_fp16, x = input_225_cast_fp16)[name = tensor("input_227_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(86562048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(96392512))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(96392704))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(96400448))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_62_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized, x = input_227_cast_fp16)[name = tensor("linear_62_cast_fp16")]; + tensor var_1263_split_sizes_0 = const()[name = tensor("op_1263_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1263_axis_0 = const()[name = tensor("op_1263_axis_0"), val = tensor(-1)]; + tensor var_1263_cast_fp16_0, tensor var_1263_cast_fp16_1 = split(axis = var_1263_axis_0, split_sizes = var_1263_split_sizes_0, x = linear_62_cast_fp16)[name = tensor("op_1263_cast_fp16")]; + tensor var_1265_mode_0 = const()[name = tensor("op_1265_mode_0"), val = tensor("EXACT")]; + tensor var_1265_cast_fp16 = gelu(mode = var_1265_mode_0, x = var_1263_cast_fp16_1)[name = tensor("op_1265_cast_fp16")]; + tensor input_229_cast_fp16 = mul(x = var_1263_cast_fp16_0, y = var_1265_cast_fp16)[name = tensor("input_229_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(96400640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(101315904))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(101316096)))]; + tensor linear_63_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized, x = input_229_cast_fp16)[name = tensor("linear_63_cast_fp16")]; + tensor hidden_states_145_cast_fp16 = add(x = linear_63_cast_fp16, y = input_225_cast_fp16)[name = tensor("hidden_states_145_cast_fp16")]; + tensor hidden_states_147_axes_0 = const()[name = tensor("hidden_states_147_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(101318720)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(101321344)))]; + tensor hidden_states_147_cast_fp16 = layer_norm(axes = hidden_states_147_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_1_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_1_norm1_weight_to_fp16, x = hidden_states_145_cast_fp16)[name = tensor("hidden_states_147_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(101323968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(102552832))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_64_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_147_cast_fp16)[name = tensor("linear_64_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(102553024))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(103781888))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_65_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_147_cast_fp16)[name = tensor("linear_65_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(103782080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(105010944))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_66_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_147_cast_fp16)[name = tensor("linear_66_cast_fp16")]; + tensor var_1300 = const()[name = tensor("op_1300"), val = tensor([1, -1, 20, 64])]; + tensor var_1301_cast_fp16 = reshape(shape = var_1300, x = linear_64_cast_fp16)[name = tensor("op_1301_cast_fp16")]; + tensor var_1303 = const()[name = tensor("op_1303"), val = tensor([1, -1, 20, 64])]; + tensor var_1304_cast_fp16 = reshape(shape = var_1303, x = linear_65_cast_fp16)[name = tensor("op_1304_cast_fp16")]; + tensor var_1306 = const()[name = tensor("op_1306"), val = tensor([1, -1, 20, 64])]; + tensor var_1307_cast_fp16 = reshape(shape = var_1306, x = linear_66_cast_fp16)[name = tensor("op_1307_cast_fp16")]; + tensor value_43_perm_0 = const()[name = tensor("value_43_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_23_y_0_to_fp16 = const()[name = tensor("mul_23_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_23_cast_fp16 = mul(x = var_1301_cast_fp16, y = mul_23_y_0_to_fp16)[name = tensor("mul_23_cast_fp16")]; + tensor matmul_10_transpose_y_0 = const()[name = tensor("matmul_10_transpose_y_0"), val = tensor(true)]; + tensor matmul_10_transpose_x_0 = const()[name = tensor("matmul_10_transpose_x_0"), val = tensor(false)]; + tensor transpose_292_perm_0 = const()[name = tensor("transpose_292_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_293_perm_0 = const()[name = tensor("transpose_293_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_643 = transpose(perm = transpose_293_perm_0, x = var_1304_cast_fp16)[name = tensor("transpose_643")]; + tensor transpose_644 = transpose(perm = transpose_292_perm_0, x = mul_23_cast_fp16)[name = tensor("transpose_644")]; + tensor matmul_10_cast_fp16 = matmul(transpose_x = matmul_10_transpose_x_0, transpose_y = matmul_10_transpose_y_0, x = transpose_644, y = transpose_643)[name = tensor("matmul_10_cast_fp16")]; + tensor softmax_10_axis_0 = const()[name = tensor("softmax_10_axis_0"), val = tensor(-1)]; + tensor softmax_10_cast_fp16 = softmax(axis = softmax_10_axis_0, x = matmul_10_cast_fp16)[name = tensor("softmax_10_cast_fp16")]; + tensor hidden_states_149_transpose_x_0 = const()[name = tensor("hidden_states_149_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_149_transpose_y_0 = const()[name = tensor("hidden_states_149_transpose_y_0"), val = tensor(false)]; + tensor transpose_642 = transpose(perm = value_43_perm_0, x = var_1307_cast_fp16)[name = tensor("transpose_642")]; + tensor hidden_states_149_cast_fp16 = matmul(transpose_x = hidden_states_149_transpose_x_0, transpose_y = hidden_states_149_transpose_y_0, x = softmax_10_cast_fp16, y = transpose_642)[name = tensor("hidden_states_149_cast_fp16")]; + tensor var_1310_perm_0 = const()[name = tensor("op_1310_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1314 = const()[name = tensor("op_1314"), val = tensor([1, -1, 1280])]; + tensor transpose_641 = transpose(perm = var_1310_perm_0, x = hidden_states_149_cast_fp16)[name = tensor("transpose_641")]; + tensor hidden_states_151_cast_fp16 = reshape(shape = var_1314, x = transpose_641)[name = tensor("hidden_states_151_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(105011136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(106240000))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(106240192)))]; + tensor linear_67_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_151_cast_fp16)[name = tensor("linear_67_cast_fp16")]; + tensor input_237_cast_fp16 = add(x = linear_67_cast_fp16, y = hidden_states_145_cast_fp16)[name = tensor("input_237_cast_fp16")]; + tensor input_239_axes_0 = const()[name = tensor("input_239_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(106242816)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(106245440)))]; + tensor input_239_cast_fp16 = layer_norm(axes = input_239_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_1_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_1_norm2_weight_to_fp16, x = input_237_cast_fp16)[name = tensor("input_239_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(106248064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(107476928))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_68_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized, x = input_239_cast_fp16)[name = tensor("linear_68_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(107477120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(109443264))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_69_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_69_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(109443456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(111409600))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_70_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_70_cast_fp16")]; + tensor var_1346 = const()[name = tensor("op_1346"), val = tensor([1, -1, 20, 64])]; + tensor var_1347_cast_fp16 = reshape(shape = var_1346, x = linear_68_cast_fp16)[name = tensor("op_1347_cast_fp16")]; + tensor var_1349 = const()[name = tensor("op_1349"), val = tensor([1, -1, 20, 64])]; + tensor var_1350_cast_fp16 = reshape(shape = var_1349, x = linear_69_cast_fp16)[name = tensor("op_1350_cast_fp16")]; + tensor var_1352 = const()[name = tensor("op_1352"), val = tensor([1, -1, 20, 64])]; + tensor var_1353_cast_fp16 = reshape(shape = var_1352, x = linear_70_cast_fp16)[name = tensor("op_1353_cast_fp16")]; + tensor value_47_perm_0 = const()[name = tensor("value_47_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_24_y_0_to_fp16 = const()[name = tensor("mul_24_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_24_cast_fp16 = mul(x = var_1347_cast_fp16, y = mul_24_y_0_to_fp16)[name = tensor("mul_24_cast_fp16")]; + tensor matmul_11_transpose_y_0 = const()[name = tensor("matmul_11_transpose_y_0"), val = tensor(true)]; + tensor matmul_11_transpose_x_0 = const()[name = tensor("matmul_11_transpose_x_0"), val = tensor(false)]; + tensor transpose_294_perm_0 = const()[name = tensor("transpose_294_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_295_perm_0 = const()[name = tensor("transpose_295_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_639 = transpose(perm = transpose_295_perm_0, x = var_1350_cast_fp16)[name = tensor("transpose_639")]; + tensor transpose_640 = transpose(perm = transpose_294_perm_0, x = mul_24_cast_fp16)[name = tensor("transpose_640")]; + tensor matmul_11_cast_fp16 = matmul(transpose_x = matmul_11_transpose_x_0, transpose_y = matmul_11_transpose_y_0, x = transpose_640, y = transpose_639)[name = tensor("matmul_11_cast_fp16")]; + tensor softmax_11_axis_0 = const()[name = tensor("softmax_11_axis_0"), val = tensor(-1)]; + tensor softmax_11_cast_fp16 = softmax(axis = softmax_11_axis_0, x = matmul_11_cast_fp16)[name = tensor("softmax_11_cast_fp16")]; + tensor hidden_states_155_transpose_x_0 = const()[name = tensor("hidden_states_155_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_155_transpose_y_0 = const()[name = tensor("hidden_states_155_transpose_y_0"), val = tensor(false)]; + tensor transpose_638 = transpose(perm = value_47_perm_0, x = var_1353_cast_fp16)[name = tensor("transpose_638")]; + tensor hidden_states_155_cast_fp16 = matmul(transpose_x = hidden_states_155_transpose_x_0, transpose_y = hidden_states_155_transpose_y_0, x = softmax_11_cast_fp16, y = transpose_638)[name = tensor("hidden_states_155_cast_fp16")]; + tensor var_1356_perm_0 = const()[name = tensor("op_1356_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1360 = const()[name = tensor("op_1360"), val = tensor([1, -1, 1280])]; + tensor transpose_637 = transpose(perm = var_1356_perm_0, x = hidden_states_155_cast_fp16)[name = tensor("transpose_637")]; + tensor hidden_states_157_cast_fp16 = reshape(shape = var_1360, x = transpose_637)[name = tensor("hidden_states_157_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(111409792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(112638656))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(112638848)))]; + tensor linear_71_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_157_cast_fp16)[name = tensor("linear_71_cast_fp16")]; + tensor input_245_cast_fp16 = add(x = linear_71_cast_fp16, y = input_237_cast_fp16)[name = tensor("input_245_cast_fp16")]; + tensor input_247_axes_0 = const()[name = tensor("input_247_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(112641472)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(112644096)))]; + tensor input_247_cast_fp16 = layer_norm(axes = input_247_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_1_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_1_norm3_weight_to_fp16, x = input_245_cast_fp16)[name = tensor("input_247_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(112646720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(122477184))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(122477376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(122485120))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_72_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized, x = input_247_cast_fp16)[name = tensor("linear_72_cast_fp16")]; + tensor var_1382_split_sizes_0 = const()[name = tensor("op_1382_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1382_axis_0 = const()[name = tensor("op_1382_axis_0"), val = tensor(-1)]; + tensor var_1382_cast_fp16_0, tensor var_1382_cast_fp16_1 = split(axis = var_1382_axis_0, split_sizes = var_1382_split_sizes_0, x = linear_72_cast_fp16)[name = tensor("op_1382_cast_fp16")]; + tensor var_1384_mode_0 = const()[name = tensor("op_1384_mode_0"), val = tensor("EXACT")]; + tensor var_1384_cast_fp16 = gelu(mode = var_1384_mode_0, x = var_1382_cast_fp16_1)[name = tensor("op_1384_cast_fp16")]; + tensor input_249_cast_fp16 = mul(x = var_1382_cast_fp16_0, y = var_1384_cast_fp16)[name = tensor("input_249_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(122485312))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(127400576))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(127400768)))]; + tensor linear_73_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized, x = input_249_cast_fp16)[name = tensor("linear_73_cast_fp16")]; + tensor hidden_states_165_cast_fp16 = add(x = linear_73_cast_fp16, y = input_245_cast_fp16)[name = tensor("hidden_states_165_cast_fp16")]; + tensor hidden_states_167_axes_0 = const()[name = tensor("hidden_states_167_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(127403392)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(127406016)))]; + tensor hidden_states_167_cast_fp16 = layer_norm(axes = hidden_states_167_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_2_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_2_norm1_weight_to_fp16, x = hidden_states_165_cast_fp16)[name = tensor("hidden_states_167_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(127408640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(128637504))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_74_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_167_cast_fp16)[name = tensor("linear_74_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(128637696))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(129866560))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_75_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_167_cast_fp16)[name = tensor("linear_75_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(129866752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(131095616))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_76_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_167_cast_fp16)[name = tensor("linear_76_cast_fp16")]; + tensor var_1419 = const()[name = tensor("op_1419"), val = tensor([1, -1, 20, 64])]; + tensor var_1420_cast_fp16 = reshape(shape = var_1419, x = linear_74_cast_fp16)[name = tensor("op_1420_cast_fp16")]; + tensor var_1422 = const()[name = tensor("op_1422"), val = tensor([1, -1, 20, 64])]; + tensor var_1423_cast_fp16 = reshape(shape = var_1422, x = linear_75_cast_fp16)[name = tensor("op_1423_cast_fp16")]; + tensor var_1425 = const()[name = tensor("op_1425"), val = tensor([1, -1, 20, 64])]; + tensor var_1426_cast_fp16 = reshape(shape = var_1425, x = linear_76_cast_fp16)[name = tensor("op_1426_cast_fp16")]; + tensor value_51_perm_0 = const()[name = tensor("value_51_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_25_y_0_to_fp16 = const()[name = tensor("mul_25_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_25_cast_fp16 = mul(x = var_1420_cast_fp16, y = mul_25_y_0_to_fp16)[name = tensor("mul_25_cast_fp16")]; + tensor matmul_12_transpose_y_0 = const()[name = tensor("matmul_12_transpose_y_0"), val = tensor(true)]; + tensor matmul_12_transpose_x_0 = const()[name = tensor("matmul_12_transpose_x_0"), val = tensor(false)]; + tensor transpose_296_perm_0 = const()[name = tensor("transpose_296_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_297_perm_0 = const()[name = tensor("transpose_297_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_635 = transpose(perm = transpose_297_perm_0, x = var_1423_cast_fp16)[name = tensor("transpose_635")]; + tensor transpose_636 = transpose(perm = transpose_296_perm_0, x = mul_25_cast_fp16)[name = tensor("transpose_636")]; + tensor matmul_12_cast_fp16 = matmul(transpose_x = matmul_12_transpose_x_0, transpose_y = matmul_12_transpose_y_0, x = transpose_636, y = transpose_635)[name = tensor("matmul_12_cast_fp16")]; + tensor softmax_12_axis_0 = const()[name = tensor("softmax_12_axis_0"), val = tensor(-1)]; + tensor softmax_12_cast_fp16 = softmax(axis = softmax_12_axis_0, x = matmul_12_cast_fp16)[name = tensor("softmax_12_cast_fp16")]; + tensor hidden_states_169_transpose_x_0 = const()[name = tensor("hidden_states_169_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_169_transpose_y_0 = const()[name = tensor("hidden_states_169_transpose_y_0"), val = tensor(false)]; + tensor transpose_634 = transpose(perm = value_51_perm_0, x = var_1426_cast_fp16)[name = tensor("transpose_634")]; + tensor hidden_states_169_cast_fp16 = matmul(transpose_x = hidden_states_169_transpose_x_0, transpose_y = hidden_states_169_transpose_y_0, x = softmax_12_cast_fp16, y = transpose_634)[name = tensor("hidden_states_169_cast_fp16")]; + tensor var_1429_perm_0 = const()[name = tensor("op_1429_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1433 = const()[name = tensor("op_1433"), val = tensor([1, -1, 1280])]; + tensor transpose_633 = transpose(perm = var_1429_perm_0, x = hidden_states_169_cast_fp16)[name = tensor("transpose_633")]; + tensor hidden_states_171_cast_fp16 = reshape(shape = var_1433, x = transpose_633)[name = tensor("hidden_states_171_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(131095808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(132324672))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(132324864)))]; + tensor linear_77_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_171_cast_fp16)[name = tensor("linear_77_cast_fp16")]; + tensor input_257_cast_fp16 = add(x = linear_77_cast_fp16, y = hidden_states_165_cast_fp16)[name = tensor("input_257_cast_fp16")]; + tensor input_259_axes_0 = const()[name = tensor("input_259_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(132327488)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(132330112)))]; + tensor input_259_cast_fp16 = layer_norm(axes = input_259_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_2_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_2_norm2_weight_to_fp16, x = input_257_cast_fp16)[name = tensor("input_259_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(132332736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(133561600))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_78_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized, x = input_259_cast_fp16)[name = tensor("linear_78_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(133561792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(135527936))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_79_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_79_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(135528128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(137494272))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_80_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_80_cast_fp16")]; + tensor var_1465 = const()[name = tensor("op_1465"), val = tensor([1, -1, 20, 64])]; + tensor var_1466_cast_fp16 = reshape(shape = var_1465, x = linear_78_cast_fp16)[name = tensor("op_1466_cast_fp16")]; + tensor var_1468 = const()[name = tensor("op_1468"), val = tensor([1, -1, 20, 64])]; + tensor var_1469_cast_fp16 = reshape(shape = var_1468, x = linear_79_cast_fp16)[name = tensor("op_1469_cast_fp16")]; + tensor var_1471 = const()[name = tensor("op_1471"), val = tensor([1, -1, 20, 64])]; + tensor var_1472_cast_fp16 = reshape(shape = var_1471, x = linear_80_cast_fp16)[name = tensor("op_1472_cast_fp16")]; + tensor value_55_perm_0 = const()[name = tensor("value_55_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_26_y_0_to_fp16 = const()[name = tensor("mul_26_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_26_cast_fp16 = mul(x = var_1466_cast_fp16, y = mul_26_y_0_to_fp16)[name = tensor("mul_26_cast_fp16")]; + tensor matmul_13_transpose_y_0 = const()[name = tensor("matmul_13_transpose_y_0"), val = tensor(true)]; + tensor matmul_13_transpose_x_0 = const()[name = tensor("matmul_13_transpose_x_0"), val = tensor(false)]; + tensor transpose_298_perm_0 = const()[name = tensor("transpose_298_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_299_perm_0 = const()[name = tensor("transpose_299_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_631 = transpose(perm = transpose_299_perm_0, x = var_1469_cast_fp16)[name = tensor("transpose_631")]; + tensor transpose_632 = transpose(perm = transpose_298_perm_0, x = mul_26_cast_fp16)[name = tensor("transpose_632")]; + tensor matmul_13_cast_fp16 = matmul(transpose_x = matmul_13_transpose_x_0, transpose_y = matmul_13_transpose_y_0, x = transpose_632, y = transpose_631)[name = tensor("matmul_13_cast_fp16")]; + tensor softmax_13_axis_0 = const()[name = tensor("softmax_13_axis_0"), val = tensor(-1)]; + tensor softmax_13_cast_fp16 = softmax(axis = softmax_13_axis_0, x = matmul_13_cast_fp16)[name = tensor("softmax_13_cast_fp16")]; + tensor hidden_states_175_transpose_x_0 = const()[name = tensor("hidden_states_175_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_175_transpose_y_0 = const()[name = tensor("hidden_states_175_transpose_y_0"), val = tensor(false)]; + tensor transpose_630 = transpose(perm = value_55_perm_0, x = var_1472_cast_fp16)[name = tensor("transpose_630")]; + tensor hidden_states_175_cast_fp16 = matmul(transpose_x = hidden_states_175_transpose_x_0, transpose_y = hidden_states_175_transpose_y_0, x = softmax_13_cast_fp16, y = transpose_630)[name = tensor("hidden_states_175_cast_fp16")]; + tensor var_1475_perm_0 = const()[name = tensor("op_1475_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1479 = const()[name = tensor("op_1479"), val = tensor([1, -1, 1280])]; + tensor transpose_629 = transpose(perm = var_1475_perm_0, x = hidden_states_175_cast_fp16)[name = tensor("transpose_629")]; + tensor hidden_states_177_cast_fp16 = reshape(shape = var_1479, x = transpose_629)[name = tensor("hidden_states_177_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(137494464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(138723328))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(138723520)))]; + tensor linear_81_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_177_cast_fp16)[name = tensor("linear_81_cast_fp16")]; + tensor input_265_cast_fp16 = add(x = linear_81_cast_fp16, y = input_257_cast_fp16)[name = tensor("input_265_cast_fp16")]; + tensor input_267_axes_0 = const()[name = tensor("input_267_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(138726144)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(138728768)))]; + tensor input_267_cast_fp16 = layer_norm(axes = input_267_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_2_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_2_norm3_weight_to_fp16, x = input_265_cast_fp16)[name = tensor("input_267_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(138731392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(148561856))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(148562048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(148569792))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_82_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized, x = input_267_cast_fp16)[name = tensor("linear_82_cast_fp16")]; + tensor var_1501_split_sizes_0 = const()[name = tensor("op_1501_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1501_axis_0 = const()[name = tensor("op_1501_axis_0"), val = tensor(-1)]; + tensor var_1501_cast_fp16_0, tensor var_1501_cast_fp16_1 = split(axis = var_1501_axis_0, split_sizes = var_1501_split_sizes_0, x = linear_82_cast_fp16)[name = tensor("op_1501_cast_fp16")]; + tensor var_1503_mode_0 = const()[name = tensor("op_1503_mode_0"), val = tensor("EXACT")]; + tensor var_1503_cast_fp16 = gelu(mode = var_1503_mode_0, x = var_1501_cast_fp16_1)[name = tensor("op_1503_cast_fp16")]; + tensor input_269_cast_fp16 = mul(x = var_1501_cast_fp16_0, y = var_1503_cast_fp16)[name = tensor("input_269_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(148569984))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153485248))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153485440)))]; + tensor linear_83_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized, x = input_269_cast_fp16)[name = tensor("linear_83_cast_fp16")]; + tensor hidden_states_185_cast_fp16 = add(x = linear_83_cast_fp16, y = input_265_cast_fp16)[name = tensor("hidden_states_185_cast_fp16")]; + tensor hidden_states_187_axes_0 = const()[name = tensor("hidden_states_187_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153488064)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153490688)))]; + tensor hidden_states_187_cast_fp16 = layer_norm(axes = hidden_states_187_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_3_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_3_norm1_weight_to_fp16, x = hidden_states_185_cast_fp16)[name = tensor("hidden_states_187_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(153493312))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(154722176))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_84_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_187_cast_fp16)[name = tensor("linear_84_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(154722368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(155951232))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_85_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_187_cast_fp16)[name = tensor("linear_85_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(155951424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(157180288))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_86_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_187_cast_fp16)[name = tensor("linear_86_cast_fp16")]; + tensor var_1538 = const()[name = tensor("op_1538"), val = tensor([1, -1, 20, 64])]; + tensor var_1539_cast_fp16 = reshape(shape = var_1538, x = linear_84_cast_fp16)[name = tensor("op_1539_cast_fp16")]; + tensor var_1541 = const()[name = tensor("op_1541"), val = tensor([1, -1, 20, 64])]; + tensor var_1542_cast_fp16 = reshape(shape = var_1541, x = linear_85_cast_fp16)[name = tensor("op_1542_cast_fp16")]; + tensor var_1544 = const()[name = tensor("op_1544"), val = tensor([1, -1, 20, 64])]; + tensor var_1545_cast_fp16 = reshape(shape = var_1544, x = linear_86_cast_fp16)[name = tensor("op_1545_cast_fp16")]; + tensor value_59_perm_0 = const()[name = tensor("value_59_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_27_y_0_to_fp16 = const()[name = tensor("mul_27_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_27_cast_fp16 = mul(x = var_1539_cast_fp16, y = mul_27_y_0_to_fp16)[name = tensor("mul_27_cast_fp16")]; + tensor matmul_14_transpose_y_0 = const()[name = tensor("matmul_14_transpose_y_0"), val = tensor(true)]; + tensor matmul_14_transpose_x_0 = const()[name = tensor("matmul_14_transpose_x_0"), val = tensor(false)]; + tensor transpose_300_perm_0 = const()[name = tensor("transpose_300_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_301_perm_0 = const()[name = tensor("transpose_301_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_627 = transpose(perm = transpose_301_perm_0, x = var_1542_cast_fp16)[name = tensor("transpose_627")]; + tensor transpose_628 = transpose(perm = transpose_300_perm_0, x = mul_27_cast_fp16)[name = tensor("transpose_628")]; + tensor matmul_14_cast_fp16 = matmul(transpose_x = matmul_14_transpose_x_0, transpose_y = matmul_14_transpose_y_0, x = transpose_628, y = transpose_627)[name = tensor("matmul_14_cast_fp16")]; + tensor softmax_14_axis_0 = const()[name = tensor("softmax_14_axis_0"), val = tensor(-1)]; + tensor softmax_14_cast_fp16 = softmax(axis = softmax_14_axis_0, x = matmul_14_cast_fp16)[name = tensor("softmax_14_cast_fp16")]; + tensor hidden_states_189_transpose_x_0 = const()[name = tensor("hidden_states_189_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_189_transpose_y_0 = const()[name = tensor("hidden_states_189_transpose_y_0"), val = tensor(false)]; + tensor transpose_626 = transpose(perm = value_59_perm_0, x = var_1545_cast_fp16)[name = tensor("transpose_626")]; + tensor hidden_states_189_cast_fp16 = matmul(transpose_x = hidden_states_189_transpose_x_0, transpose_y = hidden_states_189_transpose_y_0, x = softmax_14_cast_fp16, y = transpose_626)[name = tensor("hidden_states_189_cast_fp16")]; + tensor var_1548_perm_0 = const()[name = tensor("op_1548_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1552 = const()[name = tensor("op_1552"), val = tensor([1, -1, 1280])]; + tensor transpose_625 = transpose(perm = var_1548_perm_0, x = hidden_states_189_cast_fp16)[name = tensor("transpose_625")]; + tensor hidden_states_191_cast_fp16 = reshape(shape = var_1552, x = transpose_625)[name = tensor("hidden_states_191_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(157180480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(158409344))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(158409536)))]; + tensor linear_87_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_191_cast_fp16)[name = tensor("linear_87_cast_fp16")]; + tensor input_277_cast_fp16 = add(x = linear_87_cast_fp16, y = hidden_states_185_cast_fp16)[name = tensor("input_277_cast_fp16")]; + tensor input_279_axes_0 = const()[name = tensor("input_279_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(158412160)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(158414784)))]; + tensor input_279_cast_fp16 = layer_norm(axes = input_279_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_3_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_3_norm2_weight_to_fp16, x = input_277_cast_fp16)[name = tensor("input_279_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(158417408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(159646272))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_88_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized, x = input_279_cast_fp16)[name = tensor("linear_88_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(159646464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(161612608))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_89_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_89_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(161612800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(163578944))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_90_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_90_cast_fp16")]; + tensor var_1584 = const()[name = tensor("op_1584"), val = tensor([1, -1, 20, 64])]; + tensor var_1585_cast_fp16 = reshape(shape = var_1584, x = linear_88_cast_fp16)[name = tensor("op_1585_cast_fp16")]; + tensor var_1587 = const()[name = tensor("op_1587"), val = tensor([1, -1, 20, 64])]; + tensor var_1588_cast_fp16 = reshape(shape = var_1587, x = linear_89_cast_fp16)[name = tensor("op_1588_cast_fp16")]; + tensor var_1590 = const()[name = tensor("op_1590"), val = tensor([1, -1, 20, 64])]; + tensor var_1591_cast_fp16 = reshape(shape = var_1590, x = linear_90_cast_fp16)[name = tensor("op_1591_cast_fp16")]; + tensor value_63_perm_0 = const()[name = tensor("value_63_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_28_y_0_to_fp16 = const()[name = tensor("mul_28_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_28_cast_fp16 = mul(x = var_1585_cast_fp16, y = mul_28_y_0_to_fp16)[name = tensor("mul_28_cast_fp16")]; + tensor matmul_15_transpose_y_0 = const()[name = tensor("matmul_15_transpose_y_0"), val = tensor(true)]; + tensor matmul_15_transpose_x_0 = const()[name = tensor("matmul_15_transpose_x_0"), val = tensor(false)]; + tensor transpose_302_perm_0 = const()[name = tensor("transpose_302_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_303_perm_0 = const()[name = tensor("transpose_303_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_623 = transpose(perm = transpose_303_perm_0, x = var_1588_cast_fp16)[name = tensor("transpose_623")]; + tensor transpose_624 = transpose(perm = transpose_302_perm_0, x = mul_28_cast_fp16)[name = tensor("transpose_624")]; + tensor matmul_15_cast_fp16 = matmul(transpose_x = matmul_15_transpose_x_0, transpose_y = matmul_15_transpose_y_0, x = transpose_624, y = transpose_623)[name = tensor("matmul_15_cast_fp16")]; + tensor softmax_15_axis_0 = const()[name = tensor("softmax_15_axis_0"), val = tensor(-1)]; + tensor softmax_15_cast_fp16 = softmax(axis = softmax_15_axis_0, x = matmul_15_cast_fp16)[name = tensor("softmax_15_cast_fp16")]; + tensor hidden_states_195_transpose_x_0 = const()[name = tensor("hidden_states_195_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_195_transpose_y_0 = const()[name = tensor("hidden_states_195_transpose_y_0"), val = tensor(false)]; + tensor transpose_622 = transpose(perm = value_63_perm_0, x = var_1591_cast_fp16)[name = tensor("transpose_622")]; + tensor hidden_states_195_cast_fp16 = matmul(transpose_x = hidden_states_195_transpose_x_0, transpose_y = hidden_states_195_transpose_y_0, x = softmax_15_cast_fp16, y = transpose_622)[name = tensor("hidden_states_195_cast_fp16")]; + tensor var_1594_perm_0 = const()[name = tensor("op_1594_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1598 = const()[name = tensor("op_1598"), val = tensor([1, -1, 1280])]; + tensor transpose_621 = transpose(perm = var_1594_perm_0, x = hidden_states_195_cast_fp16)[name = tensor("transpose_621")]; + tensor hidden_states_197_cast_fp16 = reshape(shape = var_1598, x = transpose_621)[name = tensor("hidden_states_197_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(163579136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(164808000))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(164808192)))]; + tensor linear_91_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_197_cast_fp16)[name = tensor("linear_91_cast_fp16")]; + tensor input_285_cast_fp16 = add(x = linear_91_cast_fp16, y = input_277_cast_fp16)[name = tensor("input_285_cast_fp16")]; + tensor input_287_axes_0 = const()[name = tensor("input_287_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(164810816)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(164813440)))]; + tensor input_287_cast_fp16 = layer_norm(axes = input_287_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_3_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_3_norm3_weight_to_fp16, x = input_285_cast_fp16)[name = tensor("input_287_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(164816064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(174646528))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(174646720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(174654464))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_92_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized, x = input_287_cast_fp16)[name = tensor("linear_92_cast_fp16")]; + tensor var_1620_split_sizes_0 = const()[name = tensor("op_1620_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1620_axis_0 = const()[name = tensor("op_1620_axis_0"), val = tensor(-1)]; + tensor var_1620_cast_fp16_0, tensor var_1620_cast_fp16_1 = split(axis = var_1620_axis_0, split_sizes = var_1620_split_sizes_0, x = linear_92_cast_fp16)[name = tensor("op_1620_cast_fp16")]; + tensor var_1622_mode_0 = const()[name = tensor("op_1622_mode_0"), val = tensor("EXACT")]; + tensor var_1622_cast_fp16 = gelu(mode = var_1622_mode_0, x = var_1620_cast_fp16_1)[name = tensor("op_1622_cast_fp16")]; + tensor input_289_cast_fp16 = mul(x = var_1620_cast_fp16_0, y = var_1622_cast_fp16)[name = tensor("input_289_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(174654656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(179569920))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(179570112)))]; + tensor linear_93_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized, x = input_289_cast_fp16)[name = tensor("linear_93_cast_fp16")]; + tensor hidden_states_205_cast_fp16 = add(x = linear_93_cast_fp16, y = input_285_cast_fp16)[name = tensor("hidden_states_205_cast_fp16")]; + tensor hidden_states_207_axes_0 = const()[name = tensor("hidden_states_207_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(179572736)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(179575360)))]; + tensor hidden_states_207_cast_fp16 = layer_norm(axes = hidden_states_207_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_4_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_4_norm1_weight_to_fp16, x = hidden_states_205_cast_fp16)[name = tensor("hidden_states_207_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(179577984))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(180806848))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_94_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_207_cast_fp16)[name = tensor("linear_94_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(180807040))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(182035904))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_95_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_207_cast_fp16)[name = tensor("linear_95_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(182036096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(183264960))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_96_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_207_cast_fp16)[name = tensor("linear_96_cast_fp16")]; + tensor var_1657 = const()[name = tensor("op_1657"), val = tensor([1, -1, 20, 64])]; + tensor var_1658_cast_fp16 = reshape(shape = var_1657, x = linear_94_cast_fp16)[name = tensor("op_1658_cast_fp16")]; + tensor var_1660 = const()[name = tensor("op_1660"), val = tensor([1, -1, 20, 64])]; + tensor var_1661_cast_fp16 = reshape(shape = var_1660, x = linear_95_cast_fp16)[name = tensor("op_1661_cast_fp16")]; + tensor var_1663 = const()[name = tensor("op_1663"), val = tensor([1, -1, 20, 64])]; + tensor var_1664_cast_fp16 = reshape(shape = var_1663, x = linear_96_cast_fp16)[name = tensor("op_1664_cast_fp16")]; + tensor value_67_perm_0 = const()[name = tensor("value_67_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_29_y_0_to_fp16 = const()[name = tensor("mul_29_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_29_cast_fp16 = mul(x = var_1658_cast_fp16, y = mul_29_y_0_to_fp16)[name = tensor("mul_29_cast_fp16")]; + tensor matmul_16_transpose_y_0 = const()[name = tensor("matmul_16_transpose_y_0"), val = tensor(true)]; + tensor matmul_16_transpose_x_0 = const()[name = tensor("matmul_16_transpose_x_0"), val = tensor(false)]; + tensor transpose_304_perm_0 = const()[name = tensor("transpose_304_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_305_perm_0 = const()[name = tensor("transpose_305_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_619 = transpose(perm = transpose_305_perm_0, x = var_1661_cast_fp16)[name = tensor("transpose_619")]; + tensor transpose_620 = transpose(perm = transpose_304_perm_0, x = mul_29_cast_fp16)[name = tensor("transpose_620")]; + tensor matmul_16_cast_fp16 = matmul(transpose_x = matmul_16_transpose_x_0, transpose_y = matmul_16_transpose_y_0, x = transpose_620, y = transpose_619)[name = tensor("matmul_16_cast_fp16")]; + tensor softmax_16_axis_0 = const()[name = tensor("softmax_16_axis_0"), val = tensor(-1)]; + tensor softmax_16_cast_fp16 = softmax(axis = softmax_16_axis_0, x = matmul_16_cast_fp16)[name = tensor("softmax_16_cast_fp16")]; + tensor hidden_states_209_transpose_x_0 = const()[name = tensor("hidden_states_209_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_209_transpose_y_0 = const()[name = tensor("hidden_states_209_transpose_y_0"), val = tensor(false)]; + tensor transpose_618 = transpose(perm = value_67_perm_0, x = var_1664_cast_fp16)[name = tensor("transpose_618")]; + tensor hidden_states_209_cast_fp16 = matmul(transpose_x = hidden_states_209_transpose_x_0, transpose_y = hidden_states_209_transpose_y_0, x = softmax_16_cast_fp16, y = transpose_618)[name = tensor("hidden_states_209_cast_fp16")]; + tensor var_1667_perm_0 = const()[name = tensor("op_1667_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1671 = const()[name = tensor("op_1671"), val = tensor([1, -1, 1280])]; + tensor transpose_617 = transpose(perm = var_1667_perm_0, x = hidden_states_209_cast_fp16)[name = tensor("transpose_617")]; + tensor hidden_states_211_cast_fp16 = reshape(shape = var_1671, x = transpose_617)[name = tensor("hidden_states_211_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(183265152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(184494016))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(184494208)))]; + tensor linear_97_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_211_cast_fp16)[name = tensor("linear_97_cast_fp16")]; + tensor input_297_cast_fp16 = add(x = linear_97_cast_fp16, y = hidden_states_205_cast_fp16)[name = tensor("input_297_cast_fp16")]; + tensor input_299_axes_0 = const()[name = tensor("input_299_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(184496832)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(184499456)))]; + tensor input_299_cast_fp16 = layer_norm(axes = input_299_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_4_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_4_norm2_weight_to_fp16, x = input_297_cast_fp16)[name = tensor("input_299_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(184502080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(185730944))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_98_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized, x = input_299_cast_fp16)[name = tensor("linear_98_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(185731136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(187697280))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_99_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_99_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(187697472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(189663616))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_100_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_100_cast_fp16")]; + tensor var_1703 = const()[name = tensor("op_1703"), val = tensor([1, -1, 20, 64])]; + tensor var_1704_cast_fp16 = reshape(shape = var_1703, x = linear_98_cast_fp16)[name = tensor("op_1704_cast_fp16")]; + tensor var_1706 = const()[name = tensor("op_1706"), val = tensor([1, -1, 20, 64])]; + tensor var_1707_cast_fp16 = reshape(shape = var_1706, x = linear_99_cast_fp16)[name = tensor("op_1707_cast_fp16")]; + tensor var_1709 = const()[name = tensor("op_1709"), val = tensor([1, -1, 20, 64])]; + tensor var_1710_cast_fp16 = reshape(shape = var_1709, x = linear_100_cast_fp16)[name = tensor("op_1710_cast_fp16")]; + tensor value_71_perm_0 = const()[name = tensor("value_71_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_30_y_0_to_fp16 = const()[name = tensor("mul_30_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_30_cast_fp16 = mul(x = var_1704_cast_fp16, y = mul_30_y_0_to_fp16)[name = tensor("mul_30_cast_fp16")]; + tensor matmul_17_transpose_y_0 = const()[name = tensor("matmul_17_transpose_y_0"), val = tensor(true)]; + tensor matmul_17_transpose_x_0 = const()[name = tensor("matmul_17_transpose_x_0"), val = tensor(false)]; + tensor transpose_306_perm_0 = const()[name = tensor("transpose_306_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_307_perm_0 = const()[name = tensor("transpose_307_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_615 = transpose(perm = transpose_307_perm_0, x = var_1707_cast_fp16)[name = tensor("transpose_615")]; + tensor transpose_616 = transpose(perm = transpose_306_perm_0, x = mul_30_cast_fp16)[name = tensor("transpose_616")]; + tensor matmul_17_cast_fp16 = matmul(transpose_x = matmul_17_transpose_x_0, transpose_y = matmul_17_transpose_y_0, x = transpose_616, y = transpose_615)[name = tensor("matmul_17_cast_fp16")]; + tensor softmax_17_axis_0 = const()[name = tensor("softmax_17_axis_0"), val = tensor(-1)]; + tensor softmax_17_cast_fp16 = softmax(axis = softmax_17_axis_0, x = matmul_17_cast_fp16)[name = tensor("softmax_17_cast_fp16")]; + tensor hidden_states_215_transpose_x_0 = const()[name = tensor("hidden_states_215_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_215_transpose_y_0 = const()[name = tensor("hidden_states_215_transpose_y_0"), val = tensor(false)]; + tensor transpose_614 = transpose(perm = value_71_perm_0, x = var_1710_cast_fp16)[name = tensor("transpose_614")]; + tensor hidden_states_215_cast_fp16 = matmul(transpose_x = hidden_states_215_transpose_x_0, transpose_y = hidden_states_215_transpose_y_0, x = softmax_17_cast_fp16, y = transpose_614)[name = tensor("hidden_states_215_cast_fp16")]; + tensor var_1713_perm_0 = const()[name = tensor("op_1713_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1717 = const()[name = tensor("op_1717"), val = tensor([1, -1, 1280])]; + tensor transpose_613 = transpose(perm = var_1713_perm_0, x = hidden_states_215_cast_fp16)[name = tensor("transpose_613")]; + tensor hidden_states_217_cast_fp16 = reshape(shape = var_1717, x = transpose_613)[name = tensor("hidden_states_217_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(189663808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(190892672))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(190892864)))]; + tensor linear_101_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_217_cast_fp16)[name = tensor("linear_101_cast_fp16")]; + tensor input_305_cast_fp16 = add(x = linear_101_cast_fp16, y = input_297_cast_fp16)[name = tensor("input_305_cast_fp16")]; + tensor input_307_axes_0 = const()[name = tensor("input_307_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(190895488)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(190898112)))]; + tensor input_307_cast_fp16 = layer_norm(axes = input_307_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_4_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_4_norm3_weight_to_fp16, x = input_305_cast_fp16)[name = tensor("input_307_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(190900736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(200731200))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(200731392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(200739136))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_102_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized, x = input_307_cast_fp16)[name = tensor("linear_102_cast_fp16")]; + tensor var_1739_split_sizes_0 = const()[name = tensor("op_1739_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1739_axis_0 = const()[name = tensor("op_1739_axis_0"), val = tensor(-1)]; + tensor var_1739_cast_fp16_0, tensor var_1739_cast_fp16_1 = split(axis = var_1739_axis_0, split_sizes = var_1739_split_sizes_0, x = linear_102_cast_fp16)[name = tensor("op_1739_cast_fp16")]; + tensor var_1741_mode_0 = const()[name = tensor("op_1741_mode_0"), val = tensor("EXACT")]; + tensor var_1741_cast_fp16 = gelu(mode = var_1741_mode_0, x = var_1739_cast_fp16_1)[name = tensor("op_1741_cast_fp16")]; + tensor input_309_cast_fp16 = mul(x = var_1739_cast_fp16_0, y = var_1741_cast_fp16)[name = tensor("input_309_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(200739328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(205654592))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(205654784)))]; + tensor linear_103_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized, x = input_309_cast_fp16)[name = tensor("linear_103_cast_fp16")]; + tensor hidden_states_225_cast_fp16 = add(x = linear_103_cast_fp16, y = input_305_cast_fp16)[name = tensor("hidden_states_225_cast_fp16")]; + tensor hidden_states_227_axes_0 = const()[name = tensor("hidden_states_227_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(205657408)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(205660032)))]; + tensor hidden_states_227_cast_fp16 = layer_norm(axes = hidden_states_227_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_5_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_5_norm1_weight_to_fp16, x = hidden_states_225_cast_fp16)[name = tensor("hidden_states_227_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(205662656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(206891520))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_104_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_227_cast_fp16)[name = tensor("linear_104_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(206891712))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(208120576))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_105_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_227_cast_fp16)[name = tensor("linear_105_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(208120768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(209349632))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_106_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_227_cast_fp16)[name = tensor("linear_106_cast_fp16")]; + tensor var_1776 = const()[name = tensor("op_1776"), val = tensor([1, -1, 20, 64])]; + tensor var_1777_cast_fp16 = reshape(shape = var_1776, x = linear_104_cast_fp16)[name = tensor("op_1777_cast_fp16")]; + tensor var_1779 = const()[name = tensor("op_1779"), val = tensor([1, -1, 20, 64])]; + tensor var_1780_cast_fp16 = reshape(shape = var_1779, x = linear_105_cast_fp16)[name = tensor("op_1780_cast_fp16")]; + tensor var_1782 = const()[name = tensor("op_1782"), val = tensor([1, -1, 20, 64])]; + tensor var_1783_cast_fp16 = reshape(shape = var_1782, x = linear_106_cast_fp16)[name = tensor("op_1783_cast_fp16")]; + tensor value_75_perm_0 = const()[name = tensor("value_75_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_31_y_0_to_fp16 = const()[name = tensor("mul_31_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_31_cast_fp16 = mul(x = var_1777_cast_fp16, y = mul_31_y_0_to_fp16)[name = tensor("mul_31_cast_fp16")]; + tensor matmul_18_transpose_y_0 = const()[name = tensor("matmul_18_transpose_y_0"), val = tensor(true)]; + tensor matmul_18_transpose_x_0 = const()[name = tensor("matmul_18_transpose_x_0"), val = tensor(false)]; + tensor transpose_308_perm_0 = const()[name = tensor("transpose_308_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_309_perm_0 = const()[name = tensor("transpose_309_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_611 = transpose(perm = transpose_309_perm_0, x = var_1780_cast_fp16)[name = tensor("transpose_611")]; + tensor transpose_612 = transpose(perm = transpose_308_perm_0, x = mul_31_cast_fp16)[name = tensor("transpose_612")]; + tensor matmul_18_cast_fp16 = matmul(transpose_x = matmul_18_transpose_x_0, transpose_y = matmul_18_transpose_y_0, x = transpose_612, y = transpose_611)[name = tensor("matmul_18_cast_fp16")]; + tensor softmax_18_axis_0 = const()[name = tensor("softmax_18_axis_0"), val = tensor(-1)]; + tensor softmax_18_cast_fp16 = softmax(axis = softmax_18_axis_0, x = matmul_18_cast_fp16)[name = tensor("softmax_18_cast_fp16")]; + tensor hidden_states_229_transpose_x_0 = const()[name = tensor("hidden_states_229_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_229_transpose_y_0 = const()[name = tensor("hidden_states_229_transpose_y_0"), val = tensor(false)]; + tensor transpose_610 = transpose(perm = value_75_perm_0, x = var_1783_cast_fp16)[name = tensor("transpose_610")]; + tensor hidden_states_229_cast_fp16 = matmul(transpose_x = hidden_states_229_transpose_x_0, transpose_y = hidden_states_229_transpose_y_0, x = softmax_18_cast_fp16, y = transpose_610)[name = tensor("hidden_states_229_cast_fp16")]; + tensor var_1786_perm_0 = const()[name = tensor("op_1786_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1790 = const()[name = tensor("op_1790"), val = tensor([1, -1, 1280])]; + tensor transpose_609 = transpose(perm = var_1786_perm_0, x = hidden_states_229_cast_fp16)[name = tensor("transpose_609")]; + tensor hidden_states_231_cast_fp16 = reshape(shape = var_1790, x = transpose_609)[name = tensor("hidden_states_231_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(209349824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(210578688))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(210578880)))]; + tensor linear_107_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_231_cast_fp16)[name = tensor("linear_107_cast_fp16")]; + tensor input_317_cast_fp16 = add(x = linear_107_cast_fp16, y = hidden_states_225_cast_fp16)[name = tensor("input_317_cast_fp16")]; + tensor input_319_axes_0 = const()[name = tensor("input_319_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(210581504)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(210584128)))]; + tensor input_319_cast_fp16 = layer_norm(axes = input_319_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_5_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_5_norm2_weight_to_fp16, x = input_317_cast_fp16)[name = tensor("input_319_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(210586752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(211815616))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_108_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized, x = input_319_cast_fp16)[name = tensor("linear_108_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(211815808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(213781952))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_109_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_109_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(213782144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(215748288))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_110_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_110_cast_fp16")]; + tensor var_1822 = const()[name = tensor("op_1822"), val = tensor([1, -1, 20, 64])]; + tensor var_1823_cast_fp16 = reshape(shape = var_1822, x = linear_108_cast_fp16)[name = tensor("op_1823_cast_fp16")]; + tensor var_1825 = const()[name = tensor("op_1825"), val = tensor([1, -1, 20, 64])]; + tensor var_1826_cast_fp16 = reshape(shape = var_1825, x = linear_109_cast_fp16)[name = tensor("op_1826_cast_fp16")]; + tensor var_1828 = const()[name = tensor("op_1828"), val = tensor([1, -1, 20, 64])]; + tensor var_1829_cast_fp16 = reshape(shape = var_1828, x = linear_110_cast_fp16)[name = tensor("op_1829_cast_fp16")]; + tensor value_79_perm_0 = const()[name = tensor("value_79_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_32_y_0_to_fp16 = const()[name = tensor("mul_32_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_32_cast_fp16 = mul(x = var_1823_cast_fp16, y = mul_32_y_0_to_fp16)[name = tensor("mul_32_cast_fp16")]; + tensor matmul_19_transpose_y_0 = const()[name = tensor("matmul_19_transpose_y_0"), val = tensor(true)]; + tensor matmul_19_transpose_x_0 = const()[name = tensor("matmul_19_transpose_x_0"), val = tensor(false)]; + tensor transpose_310_perm_0 = const()[name = tensor("transpose_310_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_311_perm_0 = const()[name = tensor("transpose_311_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_607 = transpose(perm = transpose_311_perm_0, x = var_1826_cast_fp16)[name = tensor("transpose_607")]; + tensor transpose_608 = transpose(perm = transpose_310_perm_0, x = mul_32_cast_fp16)[name = tensor("transpose_608")]; + tensor matmul_19_cast_fp16 = matmul(transpose_x = matmul_19_transpose_x_0, transpose_y = matmul_19_transpose_y_0, x = transpose_608, y = transpose_607)[name = tensor("matmul_19_cast_fp16")]; + tensor softmax_19_axis_0 = const()[name = tensor("softmax_19_axis_0"), val = tensor(-1)]; + tensor softmax_19_cast_fp16 = softmax(axis = softmax_19_axis_0, x = matmul_19_cast_fp16)[name = tensor("softmax_19_cast_fp16")]; + tensor hidden_states_235_transpose_x_0 = const()[name = tensor("hidden_states_235_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_235_transpose_y_0 = const()[name = tensor("hidden_states_235_transpose_y_0"), val = tensor(false)]; + tensor transpose_606 = transpose(perm = value_79_perm_0, x = var_1829_cast_fp16)[name = tensor("transpose_606")]; + tensor hidden_states_235_cast_fp16 = matmul(transpose_x = hidden_states_235_transpose_x_0, transpose_y = hidden_states_235_transpose_y_0, x = softmax_19_cast_fp16, y = transpose_606)[name = tensor("hidden_states_235_cast_fp16")]; + tensor var_1832_perm_0 = const()[name = tensor("op_1832_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1836 = const()[name = tensor("op_1836"), val = tensor([1, -1, 1280])]; + tensor transpose_605 = transpose(perm = var_1832_perm_0, x = hidden_states_235_cast_fp16)[name = tensor("transpose_605")]; + tensor hidden_states_237_cast_fp16 = reshape(shape = var_1836, x = transpose_605)[name = tensor("hidden_states_237_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(215748480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(216977344))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(216977536)))]; + tensor linear_111_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_237_cast_fp16)[name = tensor("linear_111_cast_fp16")]; + tensor input_325_cast_fp16 = add(x = linear_111_cast_fp16, y = input_317_cast_fp16)[name = tensor("input_325_cast_fp16")]; + tensor input_327_axes_0 = const()[name = tensor("input_327_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(216980160)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(216982784)))]; + tensor input_327_cast_fp16 = layer_norm(axes = input_327_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_5_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_5_norm3_weight_to_fp16, x = input_325_cast_fp16)[name = tensor("input_327_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(216985408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(226815872))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(226816064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(226823808))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_112_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized, x = input_327_cast_fp16)[name = tensor("linear_112_cast_fp16")]; + tensor var_1858_split_sizes_0 = const()[name = tensor("op_1858_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1858_axis_0 = const()[name = tensor("op_1858_axis_0"), val = tensor(-1)]; + tensor var_1858_cast_fp16_0, tensor var_1858_cast_fp16_1 = split(axis = var_1858_axis_0, split_sizes = var_1858_split_sizes_0, x = linear_112_cast_fp16)[name = tensor("op_1858_cast_fp16")]; + tensor var_1860_mode_0 = const()[name = tensor("op_1860_mode_0"), val = tensor("EXACT")]; + tensor var_1860_cast_fp16 = gelu(mode = var_1860_mode_0, x = var_1858_cast_fp16_1)[name = tensor("op_1860_cast_fp16")]; + tensor input_329_cast_fp16 = mul(x = var_1858_cast_fp16_0, y = var_1860_cast_fp16)[name = tensor("input_329_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(226824000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(231739264))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(231739456)))]; + tensor linear_113_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized, x = input_329_cast_fp16)[name = tensor("linear_113_cast_fp16")]; + tensor hidden_states_245_cast_fp16 = add(x = linear_113_cast_fp16, y = input_325_cast_fp16)[name = tensor("hidden_states_245_cast_fp16")]; + tensor hidden_states_247_axes_0 = const()[name = tensor("hidden_states_247_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(231742080)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(231744704)))]; + tensor hidden_states_247_cast_fp16 = layer_norm(axes = hidden_states_247_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_6_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_6_norm1_weight_to_fp16, x = hidden_states_245_cast_fp16)[name = tensor("hidden_states_247_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(231747328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(232976192))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_114_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_247_cast_fp16)[name = tensor("linear_114_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(232976384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(234205248))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_115_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_247_cast_fp16)[name = tensor("linear_115_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(234205440))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(235434304))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_116_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_247_cast_fp16)[name = tensor("linear_116_cast_fp16")]; + tensor var_1895 = const()[name = tensor("op_1895"), val = tensor([1, -1, 20, 64])]; + tensor var_1896_cast_fp16 = reshape(shape = var_1895, x = linear_114_cast_fp16)[name = tensor("op_1896_cast_fp16")]; + tensor var_1898 = const()[name = tensor("op_1898"), val = tensor([1, -1, 20, 64])]; + tensor var_1899_cast_fp16 = reshape(shape = var_1898, x = linear_115_cast_fp16)[name = tensor("op_1899_cast_fp16")]; + tensor var_1901 = const()[name = tensor("op_1901"), val = tensor([1, -1, 20, 64])]; + tensor var_1902_cast_fp16 = reshape(shape = var_1901, x = linear_116_cast_fp16)[name = tensor("op_1902_cast_fp16")]; + tensor value_83_perm_0 = const()[name = tensor("value_83_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_33_y_0_to_fp16 = const()[name = tensor("mul_33_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_33_cast_fp16 = mul(x = var_1896_cast_fp16, y = mul_33_y_0_to_fp16)[name = tensor("mul_33_cast_fp16")]; + tensor matmul_20_transpose_y_0 = const()[name = tensor("matmul_20_transpose_y_0"), val = tensor(true)]; + tensor matmul_20_transpose_x_0 = const()[name = tensor("matmul_20_transpose_x_0"), val = tensor(false)]; + tensor transpose_312_perm_0 = const()[name = tensor("transpose_312_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_313_perm_0 = const()[name = tensor("transpose_313_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_603 = transpose(perm = transpose_313_perm_0, x = var_1899_cast_fp16)[name = tensor("transpose_603")]; + tensor transpose_604 = transpose(perm = transpose_312_perm_0, x = mul_33_cast_fp16)[name = tensor("transpose_604")]; + tensor matmul_20_cast_fp16 = matmul(transpose_x = matmul_20_transpose_x_0, transpose_y = matmul_20_transpose_y_0, x = transpose_604, y = transpose_603)[name = tensor("matmul_20_cast_fp16")]; + tensor softmax_20_axis_0 = const()[name = tensor("softmax_20_axis_0"), val = tensor(-1)]; + tensor softmax_20_cast_fp16 = softmax(axis = softmax_20_axis_0, x = matmul_20_cast_fp16)[name = tensor("softmax_20_cast_fp16")]; + tensor hidden_states_249_transpose_x_0 = const()[name = tensor("hidden_states_249_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_249_transpose_y_0 = const()[name = tensor("hidden_states_249_transpose_y_0"), val = tensor(false)]; + tensor transpose_602 = transpose(perm = value_83_perm_0, x = var_1902_cast_fp16)[name = tensor("transpose_602")]; + tensor hidden_states_249_cast_fp16 = matmul(transpose_x = hidden_states_249_transpose_x_0, transpose_y = hidden_states_249_transpose_y_0, x = softmax_20_cast_fp16, y = transpose_602)[name = tensor("hidden_states_249_cast_fp16")]; + tensor var_1905_perm_0 = const()[name = tensor("op_1905_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1909 = const()[name = tensor("op_1909"), val = tensor([1, -1, 1280])]; + tensor transpose_601 = transpose(perm = var_1905_perm_0, x = hidden_states_249_cast_fp16)[name = tensor("transpose_601")]; + tensor hidden_states_251_cast_fp16 = reshape(shape = var_1909, x = transpose_601)[name = tensor("hidden_states_251_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(235434496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(236663360))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(236663552)))]; + tensor linear_117_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_251_cast_fp16)[name = tensor("linear_117_cast_fp16")]; + tensor input_337_cast_fp16 = add(x = linear_117_cast_fp16, y = hidden_states_245_cast_fp16)[name = tensor("input_337_cast_fp16")]; + tensor input_339_axes_0 = const()[name = tensor("input_339_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(236666176)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(236668800)))]; + tensor input_339_cast_fp16 = layer_norm(axes = input_339_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_6_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_6_norm2_weight_to_fp16, x = input_337_cast_fp16)[name = tensor("input_339_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(236671424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(237900288))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_118_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized, x = input_339_cast_fp16)[name = tensor("linear_118_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(237900480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(239866624))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_119_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_119_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(239866816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(241832960))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_120_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_120_cast_fp16")]; + tensor var_1941 = const()[name = tensor("op_1941"), val = tensor([1, -1, 20, 64])]; + tensor var_1942_cast_fp16 = reshape(shape = var_1941, x = linear_118_cast_fp16)[name = tensor("op_1942_cast_fp16")]; + tensor var_1944 = const()[name = tensor("op_1944"), val = tensor([1, -1, 20, 64])]; + tensor var_1945_cast_fp16 = reshape(shape = var_1944, x = linear_119_cast_fp16)[name = tensor("op_1945_cast_fp16")]; + tensor var_1947 = const()[name = tensor("op_1947"), val = tensor([1, -1, 20, 64])]; + tensor var_1948_cast_fp16 = reshape(shape = var_1947, x = linear_120_cast_fp16)[name = tensor("op_1948_cast_fp16")]; + tensor value_87_perm_0 = const()[name = tensor("value_87_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_34_y_0_to_fp16 = const()[name = tensor("mul_34_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_34_cast_fp16 = mul(x = var_1942_cast_fp16, y = mul_34_y_0_to_fp16)[name = tensor("mul_34_cast_fp16")]; + tensor matmul_21_transpose_y_0 = const()[name = tensor("matmul_21_transpose_y_0"), val = tensor(true)]; + tensor matmul_21_transpose_x_0 = const()[name = tensor("matmul_21_transpose_x_0"), val = tensor(false)]; + tensor transpose_314_perm_0 = const()[name = tensor("transpose_314_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_315_perm_0 = const()[name = tensor("transpose_315_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_599 = transpose(perm = transpose_315_perm_0, x = var_1945_cast_fp16)[name = tensor("transpose_599")]; + tensor transpose_600 = transpose(perm = transpose_314_perm_0, x = mul_34_cast_fp16)[name = tensor("transpose_600")]; + tensor matmul_21_cast_fp16 = matmul(transpose_x = matmul_21_transpose_x_0, transpose_y = matmul_21_transpose_y_0, x = transpose_600, y = transpose_599)[name = tensor("matmul_21_cast_fp16")]; + tensor softmax_21_axis_0 = const()[name = tensor("softmax_21_axis_0"), val = tensor(-1)]; + tensor softmax_21_cast_fp16 = softmax(axis = softmax_21_axis_0, x = matmul_21_cast_fp16)[name = tensor("softmax_21_cast_fp16")]; + tensor hidden_states_255_transpose_x_0 = const()[name = tensor("hidden_states_255_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_255_transpose_y_0 = const()[name = tensor("hidden_states_255_transpose_y_0"), val = tensor(false)]; + tensor transpose_598 = transpose(perm = value_87_perm_0, x = var_1948_cast_fp16)[name = tensor("transpose_598")]; + tensor hidden_states_255_cast_fp16 = matmul(transpose_x = hidden_states_255_transpose_x_0, transpose_y = hidden_states_255_transpose_y_0, x = softmax_21_cast_fp16, y = transpose_598)[name = tensor("hidden_states_255_cast_fp16")]; + tensor var_1951_perm_0 = const()[name = tensor("op_1951_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_1955 = const()[name = tensor("op_1955"), val = tensor([1, -1, 1280])]; + tensor transpose_597 = transpose(perm = var_1951_perm_0, x = hidden_states_255_cast_fp16)[name = tensor("transpose_597")]; + tensor hidden_states_257_cast_fp16 = reshape(shape = var_1955, x = transpose_597)[name = tensor("hidden_states_257_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(241833152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(243062016))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(243062208)))]; + tensor linear_121_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_257_cast_fp16)[name = tensor("linear_121_cast_fp16")]; + tensor input_345_cast_fp16 = add(x = linear_121_cast_fp16, y = input_337_cast_fp16)[name = tensor("input_345_cast_fp16")]; + tensor input_347_axes_0 = const()[name = tensor("input_347_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(243064832)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(243067456)))]; + tensor input_347_cast_fp16 = layer_norm(axes = input_347_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_6_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_6_norm3_weight_to_fp16, x = input_345_cast_fp16)[name = tensor("input_347_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(243070080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(252900544))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(252900736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(252908480))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_122_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized, x = input_347_cast_fp16)[name = tensor("linear_122_cast_fp16")]; + tensor var_1977_split_sizes_0 = const()[name = tensor("op_1977_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_1977_axis_0 = const()[name = tensor("op_1977_axis_0"), val = tensor(-1)]; + tensor var_1977_cast_fp16_0, tensor var_1977_cast_fp16_1 = split(axis = var_1977_axis_0, split_sizes = var_1977_split_sizes_0, x = linear_122_cast_fp16)[name = tensor("op_1977_cast_fp16")]; + tensor var_1979_mode_0 = const()[name = tensor("op_1979_mode_0"), val = tensor("EXACT")]; + tensor var_1979_cast_fp16 = gelu(mode = var_1979_mode_0, x = var_1977_cast_fp16_1)[name = tensor("op_1979_cast_fp16")]; + tensor input_349_cast_fp16 = mul(x = var_1977_cast_fp16_0, y = var_1979_cast_fp16)[name = tensor("input_349_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(252908672))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(257823936))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(257824128)))]; + tensor linear_123_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized, x = input_349_cast_fp16)[name = tensor("linear_123_cast_fp16")]; + tensor hidden_states_265_cast_fp16 = add(x = linear_123_cast_fp16, y = input_345_cast_fp16)[name = tensor("hidden_states_265_cast_fp16")]; + tensor hidden_states_267_axes_0 = const()[name = tensor("hidden_states_267_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(257826752)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(257829376)))]; + tensor hidden_states_267_cast_fp16 = layer_norm(axes = hidden_states_267_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_7_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_7_norm1_weight_to_fp16, x = hidden_states_265_cast_fp16)[name = tensor("hidden_states_267_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(257832000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(259060864))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_124_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_267_cast_fp16)[name = tensor("linear_124_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(259061056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(260289920))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_125_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_267_cast_fp16)[name = tensor("linear_125_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(260290112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(261518976))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_126_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_267_cast_fp16)[name = tensor("linear_126_cast_fp16")]; + tensor var_2014 = const()[name = tensor("op_2014"), val = tensor([1, -1, 20, 64])]; + tensor var_2015_cast_fp16 = reshape(shape = var_2014, x = linear_124_cast_fp16)[name = tensor("op_2015_cast_fp16")]; + tensor var_2017 = const()[name = tensor("op_2017"), val = tensor([1, -1, 20, 64])]; + tensor var_2018_cast_fp16 = reshape(shape = var_2017, x = linear_125_cast_fp16)[name = tensor("op_2018_cast_fp16")]; + tensor var_2020 = const()[name = tensor("op_2020"), val = tensor([1, -1, 20, 64])]; + tensor var_2021_cast_fp16 = reshape(shape = var_2020, x = linear_126_cast_fp16)[name = tensor("op_2021_cast_fp16")]; + tensor value_91_perm_0 = const()[name = tensor("value_91_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_35_y_0_to_fp16 = const()[name = tensor("mul_35_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_35_cast_fp16 = mul(x = var_2015_cast_fp16, y = mul_35_y_0_to_fp16)[name = tensor("mul_35_cast_fp16")]; + tensor matmul_22_transpose_y_0 = const()[name = tensor("matmul_22_transpose_y_0"), val = tensor(true)]; + tensor matmul_22_transpose_x_0 = const()[name = tensor("matmul_22_transpose_x_0"), val = tensor(false)]; + tensor transpose_316_perm_0 = const()[name = tensor("transpose_316_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_317_perm_0 = const()[name = tensor("transpose_317_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_595 = transpose(perm = transpose_317_perm_0, x = var_2018_cast_fp16)[name = tensor("transpose_595")]; + tensor transpose_596 = transpose(perm = transpose_316_perm_0, x = mul_35_cast_fp16)[name = tensor("transpose_596")]; + tensor matmul_22_cast_fp16 = matmul(transpose_x = matmul_22_transpose_x_0, transpose_y = matmul_22_transpose_y_0, x = transpose_596, y = transpose_595)[name = tensor("matmul_22_cast_fp16")]; + tensor softmax_22_axis_0 = const()[name = tensor("softmax_22_axis_0"), val = tensor(-1)]; + tensor softmax_22_cast_fp16 = softmax(axis = softmax_22_axis_0, x = matmul_22_cast_fp16)[name = tensor("softmax_22_cast_fp16")]; + tensor hidden_states_269_transpose_x_0 = const()[name = tensor("hidden_states_269_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_269_transpose_y_0 = const()[name = tensor("hidden_states_269_transpose_y_0"), val = tensor(false)]; + tensor transpose_594 = transpose(perm = value_91_perm_0, x = var_2021_cast_fp16)[name = tensor("transpose_594")]; + tensor hidden_states_269_cast_fp16 = matmul(transpose_x = hidden_states_269_transpose_x_0, transpose_y = hidden_states_269_transpose_y_0, x = softmax_22_cast_fp16, y = transpose_594)[name = tensor("hidden_states_269_cast_fp16")]; + tensor var_2024_perm_0 = const()[name = tensor("op_2024_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2028 = const()[name = tensor("op_2028"), val = tensor([1, -1, 1280])]; + tensor transpose_593 = transpose(perm = var_2024_perm_0, x = hidden_states_269_cast_fp16)[name = tensor("transpose_593")]; + tensor hidden_states_271_cast_fp16 = reshape(shape = var_2028, x = transpose_593)[name = tensor("hidden_states_271_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(261519168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(262748032))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(262748224)))]; + tensor linear_127_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_271_cast_fp16)[name = tensor("linear_127_cast_fp16")]; + tensor input_357_cast_fp16 = add(x = linear_127_cast_fp16, y = hidden_states_265_cast_fp16)[name = tensor("input_357_cast_fp16")]; + tensor input_359_axes_0 = const()[name = tensor("input_359_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(262750848)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(262753472)))]; + tensor input_359_cast_fp16 = layer_norm(axes = input_359_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_7_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_7_norm2_weight_to_fp16, x = input_357_cast_fp16)[name = tensor("input_359_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(262756096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(263984960))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_128_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized, x = input_359_cast_fp16)[name = tensor("linear_128_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(263985152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(265951296))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_129_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_129_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(265951488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(267917632))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_130_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_130_cast_fp16")]; + tensor var_2060 = const()[name = tensor("op_2060"), val = tensor([1, -1, 20, 64])]; + tensor var_2061_cast_fp16 = reshape(shape = var_2060, x = linear_128_cast_fp16)[name = tensor("op_2061_cast_fp16")]; + tensor var_2063 = const()[name = tensor("op_2063"), val = tensor([1, -1, 20, 64])]; + tensor var_2064_cast_fp16 = reshape(shape = var_2063, x = linear_129_cast_fp16)[name = tensor("op_2064_cast_fp16")]; + tensor var_2066 = const()[name = tensor("op_2066"), val = tensor([1, -1, 20, 64])]; + tensor var_2067_cast_fp16 = reshape(shape = var_2066, x = linear_130_cast_fp16)[name = tensor("op_2067_cast_fp16")]; + tensor value_95_perm_0 = const()[name = tensor("value_95_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_36_y_0_to_fp16 = const()[name = tensor("mul_36_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_36_cast_fp16 = mul(x = var_2061_cast_fp16, y = mul_36_y_0_to_fp16)[name = tensor("mul_36_cast_fp16")]; + tensor matmul_23_transpose_y_0 = const()[name = tensor("matmul_23_transpose_y_0"), val = tensor(true)]; + tensor matmul_23_transpose_x_0 = const()[name = tensor("matmul_23_transpose_x_0"), val = tensor(false)]; + tensor transpose_318_perm_0 = const()[name = tensor("transpose_318_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_319_perm_0 = const()[name = tensor("transpose_319_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_591 = transpose(perm = transpose_319_perm_0, x = var_2064_cast_fp16)[name = tensor("transpose_591")]; + tensor transpose_592 = transpose(perm = transpose_318_perm_0, x = mul_36_cast_fp16)[name = tensor("transpose_592")]; + tensor matmul_23_cast_fp16 = matmul(transpose_x = matmul_23_transpose_x_0, transpose_y = matmul_23_transpose_y_0, x = transpose_592, y = transpose_591)[name = tensor("matmul_23_cast_fp16")]; + tensor softmax_23_axis_0 = const()[name = tensor("softmax_23_axis_0"), val = tensor(-1)]; + tensor softmax_23_cast_fp16 = softmax(axis = softmax_23_axis_0, x = matmul_23_cast_fp16)[name = tensor("softmax_23_cast_fp16")]; + tensor hidden_states_275_transpose_x_0 = const()[name = tensor("hidden_states_275_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_275_transpose_y_0 = const()[name = tensor("hidden_states_275_transpose_y_0"), val = tensor(false)]; + tensor transpose_590 = transpose(perm = value_95_perm_0, x = var_2067_cast_fp16)[name = tensor("transpose_590")]; + tensor hidden_states_275_cast_fp16 = matmul(transpose_x = hidden_states_275_transpose_x_0, transpose_y = hidden_states_275_transpose_y_0, x = softmax_23_cast_fp16, y = transpose_590)[name = tensor("hidden_states_275_cast_fp16")]; + tensor var_2070_perm_0 = const()[name = tensor("op_2070_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2074 = const()[name = tensor("op_2074"), val = tensor([1, -1, 1280])]; + tensor transpose_589 = transpose(perm = var_2070_perm_0, x = hidden_states_275_cast_fp16)[name = tensor("transpose_589")]; + tensor hidden_states_277_cast_fp16 = reshape(shape = var_2074, x = transpose_589)[name = tensor("hidden_states_277_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(267917824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(269146688))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(269146880)))]; + tensor linear_131_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_277_cast_fp16)[name = tensor("linear_131_cast_fp16")]; + tensor input_365_cast_fp16 = add(x = linear_131_cast_fp16, y = input_357_cast_fp16)[name = tensor("input_365_cast_fp16")]; + tensor input_367_axes_0 = const()[name = tensor("input_367_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(269149504)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(269152128)))]; + tensor input_367_cast_fp16 = layer_norm(axes = input_367_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_7_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_7_norm3_weight_to_fp16, x = input_365_cast_fp16)[name = tensor("input_367_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(269154752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(278985216))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(278985408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(278993152))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_132_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized, x = input_367_cast_fp16)[name = tensor("linear_132_cast_fp16")]; + tensor var_2096_split_sizes_0 = const()[name = tensor("op_2096_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2096_axis_0 = const()[name = tensor("op_2096_axis_0"), val = tensor(-1)]; + tensor var_2096_cast_fp16_0, tensor var_2096_cast_fp16_1 = split(axis = var_2096_axis_0, split_sizes = var_2096_split_sizes_0, x = linear_132_cast_fp16)[name = tensor("op_2096_cast_fp16")]; + tensor var_2098_mode_0 = const()[name = tensor("op_2098_mode_0"), val = tensor("EXACT")]; + tensor var_2098_cast_fp16 = gelu(mode = var_2098_mode_0, x = var_2096_cast_fp16_1)[name = tensor("op_2098_cast_fp16")]; + tensor input_369_cast_fp16 = mul(x = var_2096_cast_fp16_0, y = var_2098_cast_fp16)[name = tensor("input_369_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(278993344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(283908608))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(283908800)))]; + tensor linear_133_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized, x = input_369_cast_fp16)[name = tensor("linear_133_cast_fp16")]; + tensor hidden_states_285_cast_fp16 = add(x = linear_133_cast_fp16, y = input_365_cast_fp16)[name = tensor("hidden_states_285_cast_fp16")]; + tensor hidden_states_287_axes_0 = const()[name = tensor("hidden_states_287_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(283911424)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(283914048)))]; + tensor hidden_states_287_cast_fp16 = layer_norm(axes = hidden_states_287_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_8_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_8_norm1_weight_to_fp16, x = hidden_states_285_cast_fp16)[name = tensor("hidden_states_287_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(283916672))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(285145536))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_134_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_287_cast_fp16)[name = tensor("linear_134_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(285145728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(286374592))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_135_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_287_cast_fp16)[name = tensor("linear_135_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(286374784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(287603648))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_136_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_287_cast_fp16)[name = tensor("linear_136_cast_fp16")]; + tensor var_2133 = const()[name = tensor("op_2133"), val = tensor([1, -1, 20, 64])]; + tensor var_2134_cast_fp16 = reshape(shape = var_2133, x = linear_134_cast_fp16)[name = tensor("op_2134_cast_fp16")]; + tensor var_2136 = const()[name = tensor("op_2136"), val = tensor([1, -1, 20, 64])]; + tensor var_2137_cast_fp16 = reshape(shape = var_2136, x = linear_135_cast_fp16)[name = tensor("op_2137_cast_fp16")]; + tensor var_2139 = const()[name = tensor("op_2139"), val = tensor([1, -1, 20, 64])]; + tensor var_2140_cast_fp16 = reshape(shape = var_2139, x = linear_136_cast_fp16)[name = tensor("op_2140_cast_fp16")]; + tensor value_99_perm_0 = const()[name = tensor("value_99_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_37_y_0_to_fp16 = const()[name = tensor("mul_37_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_37_cast_fp16 = mul(x = var_2134_cast_fp16, y = mul_37_y_0_to_fp16)[name = tensor("mul_37_cast_fp16")]; + tensor matmul_24_transpose_y_0 = const()[name = tensor("matmul_24_transpose_y_0"), val = tensor(true)]; + tensor matmul_24_transpose_x_0 = const()[name = tensor("matmul_24_transpose_x_0"), val = tensor(false)]; + tensor transpose_320_perm_0 = const()[name = tensor("transpose_320_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_321_perm_0 = const()[name = tensor("transpose_321_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_587 = transpose(perm = transpose_321_perm_0, x = var_2137_cast_fp16)[name = tensor("transpose_587")]; + tensor transpose_588 = transpose(perm = transpose_320_perm_0, x = mul_37_cast_fp16)[name = tensor("transpose_588")]; + tensor matmul_24_cast_fp16 = matmul(transpose_x = matmul_24_transpose_x_0, transpose_y = matmul_24_transpose_y_0, x = transpose_588, y = transpose_587)[name = tensor("matmul_24_cast_fp16")]; + tensor softmax_24_axis_0 = const()[name = tensor("softmax_24_axis_0"), val = tensor(-1)]; + tensor softmax_24_cast_fp16 = softmax(axis = softmax_24_axis_0, x = matmul_24_cast_fp16)[name = tensor("softmax_24_cast_fp16")]; + tensor hidden_states_289_transpose_x_0 = const()[name = tensor("hidden_states_289_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_289_transpose_y_0 = const()[name = tensor("hidden_states_289_transpose_y_0"), val = tensor(false)]; + tensor transpose_586 = transpose(perm = value_99_perm_0, x = var_2140_cast_fp16)[name = tensor("transpose_586")]; + tensor hidden_states_289_cast_fp16 = matmul(transpose_x = hidden_states_289_transpose_x_0, transpose_y = hidden_states_289_transpose_y_0, x = softmax_24_cast_fp16, y = transpose_586)[name = tensor("hidden_states_289_cast_fp16")]; + tensor var_2143_perm_0 = const()[name = tensor("op_2143_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2147 = const()[name = tensor("op_2147"), val = tensor([1, -1, 1280])]; + tensor transpose_585 = transpose(perm = var_2143_perm_0, x = hidden_states_289_cast_fp16)[name = tensor("transpose_585")]; + tensor hidden_states_291_cast_fp16 = reshape(shape = var_2147, x = transpose_585)[name = tensor("hidden_states_291_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(287603840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(288832704))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(288832896)))]; + tensor linear_137_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_291_cast_fp16)[name = tensor("linear_137_cast_fp16")]; + tensor input_377_cast_fp16 = add(x = linear_137_cast_fp16, y = hidden_states_285_cast_fp16)[name = tensor("input_377_cast_fp16")]; + tensor input_379_axes_0 = const()[name = tensor("input_379_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(288835520)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(288838144)))]; + tensor input_379_cast_fp16 = layer_norm(axes = input_379_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_8_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_8_norm2_weight_to_fp16, x = input_377_cast_fp16)[name = tensor("input_379_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(288840768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(290069632))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_138_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized, x = input_379_cast_fp16)[name = tensor("linear_138_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(290069824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(292035968))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_139_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_139_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(292036160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(294002304))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_140_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_140_cast_fp16")]; + tensor var_2179 = const()[name = tensor("op_2179"), val = tensor([1, -1, 20, 64])]; + tensor var_2180_cast_fp16 = reshape(shape = var_2179, x = linear_138_cast_fp16)[name = tensor("op_2180_cast_fp16")]; + tensor var_2182 = const()[name = tensor("op_2182"), val = tensor([1, -1, 20, 64])]; + tensor var_2183_cast_fp16 = reshape(shape = var_2182, x = linear_139_cast_fp16)[name = tensor("op_2183_cast_fp16")]; + tensor var_2185 = const()[name = tensor("op_2185"), val = tensor([1, -1, 20, 64])]; + tensor var_2186_cast_fp16 = reshape(shape = var_2185, x = linear_140_cast_fp16)[name = tensor("op_2186_cast_fp16")]; + tensor value_103_perm_0 = const()[name = tensor("value_103_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_38_y_0_to_fp16 = const()[name = tensor("mul_38_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_38_cast_fp16 = mul(x = var_2180_cast_fp16, y = mul_38_y_0_to_fp16)[name = tensor("mul_38_cast_fp16")]; + tensor matmul_25_transpose_y_0 = const()[name = tensor("matmul_25_transpose_y_0"), val = tensor(true)]; + tensor matmul_25_transpose_x_0 = const()[name = tensor("matmul_25_transpose_x_0"), val = tensor(false)]; + tensor transpose_322_perm_0 = const()[name = tensor("transpose_322_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_323_perm_0 = const()[name = tensor("transpose_323_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_583 = transpose(perm = transpose_323_perm_0, x = var_2183_cast_fp16)[name = tensor("transpose_583")]; + tensor transpose_584 = transpose(perm = transpose_322_perm_0, x = mul_38_cast_fp16)[name = tensor("transpose_584")]; + tensor matmul_25_cast_fp16 = matmul(transpose_x = matmul_25_transpose_x_0, transpose_y = matmul_25_transpose_y_0, x = transpose_584, y = transpose_583)[name = tensor("matmul_25_cast_fp16")]; + tensor softmax_25_axis_0 = const()[name = tensor("softmax_25_axis_0"), val = tensor(-1)]; + tensor softmax_25_cast_fp16 = softmax(axis = softmax_25_axis_0, x = matmul_25_cast_fp16)[name = tensor("softmax_25_cast_fp16")]; + tensor hidden_states_295_transpose_x_0 = const()[name = tensor("hidden_states_295_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_295_transpose_y_0 = const()[name = tensor("hidden_states_295_transpose_y_0"), val = tensor(false)]; + tensor transpose_582 = transpose(perm = value_103_perm_0, x = var_2186_cast_fp16)[name = tensor("transpose_582")]; + tensor hidden_states_295_cast_fp16 = matmul(transpose_x = hidden_states_295_transpose_x_0, transpose_y = hidden_states_295_transpose_y_0, x = softmax_25_cast_fp16, y = transpose_582)[name = tensor("hidden_states_295_cast_fp16")]; + tensor var_2189_perm_0 = const()[name = tensor("op_2189_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2193 = const()[name = tensor("op_2193"), val = tensor([1, -1, 1280])]; + tensor transpose_581 = transpose(perm = var_2189_perm_0, x = hidden_states_295_cast_fp16)[name = tensor("transpose_581")]; + tensor hidden_states_297_cast_fp16 = reshape(shape = var_2193, x = transpose_581)[name = tensor("hidden_states_297_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(294002496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(295231360))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(295231552)))]; + tensor linear_141_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_297_cast_fp16)[name = tensor("linear_141_cast_fp16")]; + tensor input_385_cast_fp16 = add(x = linear_141_cast_fp16, y = input_377_cast_fp16)[name = tensor("input_385_cast_fp16")]; + tensor input_387_axes_0 = const()[name = tensor("input_387_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(295234176)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(295236800)))]; + tensor input_387_cast_fp16 = layer_norm(axes = input_387_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_8_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_8_norm3_weight_to_fp16, x = input_385_cast_fp16)[name = tensor("input_387_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(295239424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(305069888))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(305070080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(305077824))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_142_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized, x = input_387_cast_fp16)[name = tensor("linear_142_cast_fp16")]; + tensor var_2215_split_sizes_0 = const()[name = tensor("op_2215_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2215_axis_0 = const()[name = tensor("op_2215_axis_0"), val = tensor(-1)]; + tensor var_2215_cast_fp16_0, tensor var_2215_cast_fp16_1 = split(axis = var_2215_axis_0, split_sizes = var_2215_split_sizes_0, x = linear_142_cast_fp16)[name = tensor("op_2215_cast_fp16")]; + tensor var_2217_mode_0 = const()[name = tensor("op_2217_mode_0"), val = tensor("EXACT")]; + tensor var_2217_cast_fp16 = gelu(mode = var_2217_mode_0, x = var_2215_cast_fp16_1)[name = tensor("op_2217_cast_fp16")]; + tensor input_389_cast_fp16 = mul(x = var_2215_cast_fp16_0, y = var_2217_cast_fp16)[name = tensor("input_389_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(305078016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(309993280))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(309993472)))]; + tensor linear_143_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized, x = input_389_cast_fp16)[name = tensor("linear_143_cast_fp16")]; + tensor hidden_states_305_cast_fp16 = add(x = linear_143_cast_fp16, y = input_385_cast_fp16)[name = tensor("hidden_states_305_cast_fp16")]; + tensor hidden_states_307_axes_0 = const()[name = tensor("hidden_states_307_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(309996096)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(309998720)))]; + tensor hidden_states_307_cast_fp16 = layer_norm(axes = hidden_states_307_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_9_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_9_norm1_weight_to_fp16, x = hidden_states_305_cast_fp16)[name = tensor("hidden_states_307_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(310001344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(311230208))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_144_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_307_cast_fp16)[name = tensor("linear_144_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(311230400))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(312459264))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_145_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_307_cast_fp16)[name = tensor("linear_145_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(312459456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(313688320))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_146_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_307_cast_fp16)[name = tensor("linear_146_cast_fp16")]; + tensor var_2252 = const()[name = tensor("op_2252"), val = tensor([1, -1, 20, 64])]; + tensor var_2253_cast_fp16 = reshape(shape = var_2252, x = linear_144_cast_fp16)[name = tensor("op_2253_cast_fp16")]; + tensor var_2255 = const()[name = tensor("op_2255"), val = tensor([1, -1, 20, 64])]; + tensor var_2256_cast_fp16 = reshape(shape = var_2255, x = linear_145_cast_fp16)[name = tensor("op_2256_cast_fp16")]; + tensor var_2258 = const()[name = tensor("op_2258"), val = tensor([1, -1, 20, 64])]; + tensor var_2259_cast_fp16 = reshape(shape = var_2258, x = linear_146_cast_fp16)[name = tensor("op_2259_cast_fp16")]; + tensor value_107_perm_0 = const()[name = tensor("value_107_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_39_y_0_to_fp16 = const()[name = tensor("mul_39_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_39_cast_fp16 = mul(x = var_2253_cast_fp16, y = mul_39_y_0_to_fp16)[name = tensor("mul_39_cast_fp16")]; + tensor matmul_26_transpose_y_0 = const()[name = tensor("matmul_26_transpose_y_0"), val = tensor(true)]; + tensor matmul_26_transpose_x_0 = const()[name = tensor("matmul_26_transpose_x_0"), val = tensor(false)]; + tensor transpose_324_perm_0 = const()[name = tensor("transpose_324_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_325_perm_0 = const()[name = tensor("transpose_325_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_579 = transpose(perm = transpose_325_perm_0, x = var_2256_cast_fp16)[name = tensor("transpose_579")]; + tensor transpose_580 = transpose(perm = transpose_324_perm_0, x = mul_39_cast_fp16)[name = tensor("transpose_580")]; + tensor matmul_26_cast_fp16 = matmul(transpose_x = matmul_26_transpose_x_0, transpose_y = matmul_26_transpose_y_0, x = transpose_580, y = transpose_579)[name = tensor("matmul_26_cast_fp16")]; + tensor softmax_26_axis_0 = const()[name = tensor("softmax_26_axis_0"), val = tensor(-1)]; + tensor softmax_26_cast_fp16 = softmax(axis = softmax_26_axis_0, x = matmul_26_cast_fp16)[name = tensor("softmax_26_cast_fp16")]; + tensor hidden_states_309_transpose_x_0 = const()[name = tensor("hidden_states_309_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_309_transpose_y_0 = const()[name = tensor("hidden_states_309_transpose_y_0"), val = tensor(false)]; + tensor transpose_578 = transpose(perm = value_107_perm_0, x = var_2259_cast_fp16)[name = tensor("transpose_578")]; + tensor hidden_states_309_cast_fp16 = matmul(transpose_x = hidden_states_309_transpose_x_0, transpose_y = hidden_states_309_transpose_y_0, x = softmax_26_cast_fp16, y = transpose_578)[name = tensor("hidden_states_309_cast_fp16")]; + tensor var_2262_perm_0 = const()[name = tensor("op_2262_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2266 = const()[name = tensor("op_2266"), val = tensor([1, -1, 1280])]; + tensor transpose_577 = transpose(perm = var_2262_perm_0, x = hidden_states_309_cast_fp16)[name = tensor("transpose_577")]; + tensor hidden_states_311_cast_fp16 = reshape(shape = var_2266, x = transpose_577)[name = tensor("hidden_states_311_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(313688512))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(314917376))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(314917568)))]; + tensor linear_147_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_311_cast_fp16)[name = tensor("linear_147_cast_fp16")]; + tensor input_397_cast_fp16 = add(x = linear_147_cast_fp16, y = hidden_states_305_cast_fp16)[name = tensor("input_397_cast_fp16")]; + tensor input_399_axes_0 = const()[name = tensor("input_399_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(314920192)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(314922816)))]; + tensor input_399_cast_fp16 = layer_norm(axes = input_399_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_9_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_9_norm2_weight_to_fp16, x = input_397_cast_fp16)[name = tensor("input_399_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(314925440))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(316154304))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_148_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized, x = input_399_cast_fp16)[name = tensor("linear_148_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(316154496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(318120640))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_149_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_149_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(318120832))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(320086976))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_150_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_150_cast_fp16")]; + tensor var_2298 = const()[name = tensor("op_2298"), val = tensor([1, -1, 20, 64])]; + tensor var_2299_cast_fp16 = reshape(shape = var_2298, x = linear_148_cast_fp16)[name = tensor("op_2299_cast_fp16")]; + tensor var_2301 = const()[name = tensor("op_2301"), val = tensor([1, -1, 20, 64])]; + tensor var_2302_cast_fp16 = reshape(shape = var_2301, x = linear_149_cast_fp16)[name = tensor("op_2302_cast_fp16")]; + tensor var_2304 = const()[name = tensor("op_2304"), val = tensor([1, -1, 20, 64])]; + tensor var_2305_cast_fp16 = reshape(shape = var_2304, x = linear_150_cast_fp16)[name = tensor("op_2305_cast_fp16")]; + tensor value_111_perm_0 = const()[name = tensor("value_111_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_40_y_0_to_fp16 = const()[name = tensor("mul_40_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_40_cast_fp16 = mul(x = var_2299_cast_fp16, y = mul_40_y_0_to_fp16)[name = tensor("mul_40_cast_fp16")]; + tensor matmul_27_transpose_y_0 = const()[name = tensor("matmul_27_transpose_y_0"), val = tensor(true)]; + tensor matmul_27_transpose_x_0 = const()[name = tensor("matmul_27_transpose_x_0"), val = tensor(false)]; + tensor transpose_326_perm_0 = const()[name = tensor("transpose_326_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_327_perm_0 = const()[name = tensor("transpose_327_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_575 = transpose(perm = transpose_327_perm_0, x = var_2302_cast_fp16)[name = tensor("transpose_575")]; + tensor transpose_576 = transpose(perm = transpose_326_perm_0, x = mul_40_cast_fp16)[name = tensor("transpose_576")]; + tensor matmul_27_cast_fp16 = matmul(transpose_x = matmul_27_transpose_x_0, transpose_y = matmul_27_transpose_y_0, x = transpose_576, y = transpose_575)[name = tensor("matmul_27_cast_fp16")]; + tensor softmax_27_axis_0 = const()[name = tensor("softmax_27_axis_0"), val = tensor(-1)]; + tensor softmax_27_cast_fp16 = softmax(axis = softmax_27_axis_0, x = matmul_27_cast_fp16)[name = tensor("softmax_27_cast_fp16")]; + tensor hidden_states_315_transpose_x_0 = const()[name = tensor("hidden_states_315_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_315_transpose_y_0 = const()[name = tensor("hidden_states_315_transpose_y_0"), val = tensor(false)]; + tensor transpose_574 = transpose(perm = value_111_perm_0, x = var_2305_cast_fp16)[name = tensor("transpose_574")]; + tensor hidden_states_315_cast_fp16 = matmul(transpose_x = hidden_states_315_transpose_x_0, transpose_y = hidden_states_315_transpose_y_0, x = softmax_27_cast_fp16, y = transpose_574)[name = tensor("hidden_states_315_cast_fp16")]; + tensor var_2308_perm_0 = const()[name = tensor("op_2308_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2312 = const()[name = tensor("op_2312"), val = tensor([1, -1, 1280])]; + tensor transpose_573 = transpose(perm = var_2308_perm_0, x = hidden_states_315_cast_fp16)[name = tensor("transpose_573")]; + tensor hidden_states_317_cast_fp16 = reshape(shape = var_2312, x = transpose_573)[name = tensor("hidden_states_317_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(320087168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(321316032))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(321316224)))]; + tensor linear_151_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_317_cast_fp16)[name = tensor("linear_151_cast_fp16")]; + tensor input_405_cast_fp16 = add(x = linear_151_cast_fp16, y = input_397_cast_fp16)[name = tensor("input_405_cast_fp16")]; + tensor input_407_axes_0 = const()[name = tensor("input_407_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(321318848)))]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(321321472)))]; + tensor input_407_cast_fp16 = layer_norm(axes = input_407_axes_0, beta = down_blocks_2_attentions_0_transformer_blocks_9_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_0_transformer_blocks_9_norm3_weight_to_fp16, x = input_405_cast_fp16)[name = tensor("input_407_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(321324096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(331154560))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(331154752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(331162496))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_152_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized, x = input_407_cast_fp16)[name = tensor("linear_152_cast_fp16")]; + tensor var_2334_split_sizes_0 = const()[name = tensor("op_2334_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2334_axis_0 = const()[name = tensor("op_2334_axis_0"), val = tensor(-1)]; + tensor var_2334_cast_fp16_0, tensor var_2334_cast_fp16_1 = split(axis = var_2334_axis_0, split_sizes = var_2334_split_sizes_0, x = linear_152_cast_fp16)[name = tensor("op_2334_cast_fp16")]; + tensor var_2336_mode_0 = const()[name = tensor("op_2336_mode_0"), val = tensor("EXACT")]; + tensor var_2336_cast_fp16 = gelu(mode = var_2336_mode_0, x = var_2334_cast_fp16_1)[name = tensor("op_2336_cast_fp16")]; + tensor input_409_cast_fp16 = mul(x = var_2334_cast_fp16_0, y = var_2336_cast_fp16)[name = tensor("input_409_cast_fp16")]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(331162688))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(336077952))), name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(336078144)))]; + tensor linear_153_cast_fp16 = linear(bias = down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized, x = input_409_cast_fp16)[name = tensor("linear_153_cast_fp16")]; + tensor input_413_cast_fp16 = add(x = linear_153_cast_fp16, y = input_405_cast_fp16)[name = tensor("input_413_cast_fp16")]; + tensor down_blocks_2_attentions_0_proj_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(336080768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(337309632))), name = tensor("down_blocks_2_attentions_0_proj_out_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_0_proj_out_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_0_proj_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(337309824)))]; + tensor linear_154_cast_fp16 = linear(bias = down_blocks_2_attentions_0_proj_out_bias_to_fp16, weight = down_blocks_2_attentions_0_proj_out_weight_to_fp16_palettized, x = input_413_cast_fp16)[name = tensor("linear_154_cast_fp16")]; + tensor var_2346 = const()[name = tensor("op_2346"), val = tensor([1, 32, 32, 1280])]; + tensor var_2347_cast_fp16 = reshape(shape = var_2346, x = linear_154_cast_fp16)[name = tensor("op_2347_cast_fp16")]; + tensor var_2348 = const()[name = tensor("op_2348"), val = tensor([0, 3, 1, 2])]; + tensor transpose_572 = transpose(perm = var_2348, x = var_2347_cast_fp16)[name = tensor("transpose_572")]; + tensor input_415_cast_fp16 = add(x = transpose_572, y = var_1106_cast_fp16)[name = tensor("input_415_cast_fp16")]; + tensor reshape_52_shape_0 = const()[name = tensor("reshape_52_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_52_cast_fp16 = reshape(shape = reshape_52_shape_0, x = input_415_cast_fp16)[name = tensor("reshape_52_cast_fp16")]; + tensor reduce_mean_39_axes_0 = const()[name = tensor("reduce_mean_39_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_39_keep_dims_0 = const()[name = tensor("reduce_mean_39_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_39_cast_fp16 = reduce_mean(axes = reduce_mean_39_axes_0, keep_dims = reduce_mean_39_keep_dims_0, x = reshape_52_cast_fp16)[name = tensor("reduce_mean_39_cast_fp16")]; + tensor sub_26_cast_fp16 = sub(x = reshape_52_cast_fp16, y = reduce_mean_39_cast_fp16)[name = tensor("sub_26_cast_fp16")]; + tensor square_13_cast_fp16 = square(x = sub_26_cast_fp16)[name = tensor("square_13_cast_fp16")]; + tensor reduce_mean_41_axes_0 = const()[name = tensor("reduce_mean_41_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_41_keep_dims_0 = const()[name = tensor("reduce_mean_41_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_41_cast_fp16 = reduce_mean(axes = reduce_mean_41_axes_0, keep_dims = reduce_mean_41_keep_dims_0, x = square_13_cast_fp16)[name = tensor("reduce_mean_41_cast_fp16")]; + tensor add_26_y_0_to_fp16 = const()[name = tensor("add_26_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_26_cast_fp16 = add(x = reduce_mean_41_cast_fp16, y = add_26_y_0_to_fp16)[name = tensor("add_26_cast_fp16")]; + tensor sqrt_13_cast_fp16 = sqrt(x = add_26_cast_fp16)[name = tensor("sqrt_13_cast_fp16")]; + tensor real_div_13_cast_fp16 = real_div(x = sub_26_cast_fp16, y = sqrt_13_cast_fp16)[name = tensor("real_div_13_cast_fp16")]; + tensor reshape_53_shape_0 = const()[name = tensor("reshape_53_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_53_cast_fp16 = reshape(shape = reshape_53_shape_0, x = real_div_13_cast_fp16)[name = tensor("reshape_53_cast_fp16")]; + tensor add_27_gamma_0_to_fp16 = const()[name = tensor("add_27_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(337312448)))]; + tensor add_27_beta_0_to_fp16 = const()[name = tensor("add_27_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(337315072)))]; + tensor add_27_epsilon_0_to_fp16 = const()[name = tensor("add_27_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_27_cast_fp16 = batch_norm(beta = add_27_beta_0_to_fp16, epsilon = add_27_epsilon_0_to_fp16, gamma = add_27_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_53_cast_fp16)[name = tensor("add_27_cast_fp16")]; + tensor input_419_cast_fp16 = silu(x = add_27_cast_fp16)[name = tensor("input_419_cast_fp16")]; + tensor var_2363 = const()[name = tensor("op_2363"), val = tensor([1, 1])]; + tensor var_2365 = const()[name = tensor("op_2365"), val = tensor([1, 1])]; + tensor hidden_states_329_pad_type_0 = const()[name = tensor("hidden_states_329_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_329_pad_0 = const()[name = tensor("hidden_states_329_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_2_resnets_1_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(337317696))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(348376960))), name = tensor("down_blocks_2_resnets_1_conv1_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor down_blocks_2_resnets_1_conv1_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(348377152)))]; + tensor hidden_states_329_cast_fp16 = conv(bias = down_blocks_2_resnets_1_conv1_bias_to_fp16, dilations = var_2365, groups = var_1048, pad = hidden_states_329_pad_0, pad_type = hidden_states_329_pad_type_0, strides = var_2363, weight = down_blocks_2_resnets_1_conv1_weight_to_fp16_palettized, x = input_419_cast_fp16)[name = tensor("hidden_states_329_cast_fp16")]; + tensor down_blocks_2_resnets_1_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(348379776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(349608640))), name = tensor("down_blocks_2_resnets_1_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_resnets_1_time_emb_proj_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_1_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(349608832)))]; + tensor linear_155_cast_fp16 = linear(bias = down_blocks_2_resnets_1_time_emb_proj_bias_to_fp16, weight = down_blocks_2_resnets_1_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_155_cast_fp16")]; + tensor var_2374_axes_0 = const()[name = tensor("op_2374_axes_0"), val = tensor([2])]; + tensor var_2374_cast_fp16 = expand_dims(axes = var_2374_axes_0, x = linear_155_cast_fp16)[name = tensor("op_2374_cast_fp16")]; + tensor temb_11_axes_0 = const()[name = tensor("temb_11_axes_0"), val = tensor([3])]; + tensor temb_11_cast_fp16 = expand_dims(axes = temb_11_axes_0, x = var_2374_cast_fp16)[name = tensor("temb_11_cast_fp16")]; + tensor input_423_cast_fp16 = add(x = hidden_states_329_cast_fp16, y = temb_11_cast_fp16)[name = tensor("input_423_cast_fp16")]; + tensor reshape_56_shape_0 = const()[name = tensor("reshape_56_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_56_cast_fp16 = reshape(shape = reshape_56_shape_0, x = input_423_cast_fp16)[name = tensor("reshape_56_cast_fp16")]; + tensor reduce_mean_42_axes_0 = const()[name = tensor("reduce_mean_42_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_42_keep_dims_0 = const()[name = tensor("reduce_mean_42_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_42_cast_fp16 = reduce_mean(axes = reduce_mean_42_axes_0, keep_dims = reduce_mean_42_keep_dims_0, x = reshape_56_cast_fp16)[name = tensor("reduce_mean_42_cast_fp16")]; + tensor sub_28_cast_fp16 = sub(x = reshape_56_cast_fp16, y = reduce_mean_42_cast_fp16)[name = tensor("sub_28_cast_fp16")]; + tensor square_14_cast_fp16 = square(x = sub_28_cast_fp16)[name = tensor("square_14_cast_fp16")]; + tensor reduce_mean_44_axes_0 = const()[name = tensor("reduce_mean_44_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_44_keep_dims_0 = const()[name = tensor("reduce_mean_44_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_44_cast_fp16 = reduce_mean(axes = reduce_mean_44_axes_0, keep_dims = reduce_mean_44_keep_dims_0, x = square_14_cast_fp16)[name = tensor("reduce_mean_44_cast_fp16")]; + tensor add_28_y_0_to_fp16 = const()[name = tensor("add_28_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_28_cast_fp16 = add(x = reduce_mean_44_cast_fp16, y = add_28_y_0_to_fp16)[name = tensor("add_28_cast_fp16")]; + tensor sqrt_14_cast_fp16 = sqrt(x = add_28_cast_fp16)[name = tensor("sqrt_14_cast_fp16")]; + tensor real_div_14_cast_fp16 = real_div(x = sub_28_cast_fp16, y = sqrt_14_cast_fp16)[name = tensor("real_div_14_cast_fp16")]; + tensor reshape_57_shape_0 = const()[name = tensor("reshape_57_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_57_cast_fp16 = reshape(shape = reshape_57_shape_0, x = real_div_14_cast_fp16)[name = tensor("reshape_57_cast_fp16")]; + tensor add_29_gamma_0_to_fp16 = const()[name = tensor("add_29_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(349611456)))]; + tensor add_29_beta_0_to_fp16 = const()[name = tensor("add_29_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(349614080)))]; + tensor add_29_epsilon_0_to_fp16 = const()[name = tensor("add_29_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_29_cast_fp16 = batch_norm(beta = add_29_beta_0_to_fp16, epsilon = add_29_epsilon_0_to_fp16, gamma = add_29_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_57_cast_fp16)[name = tensor("add_29_cast_fp16")]; + tensor input_427_cast_fp16 = silu(x = add_29_cast_fp16)[name = tensor("input_427_cast_fp16")]; + tensor var_2384 = const()[name = tensor("op_2384"), val = tensor([1, 1])]; + tensor var_2386 = const()[name = tensor("op_2386"), val = tensor([1, 1])]; + tensor hidden_states_331_pad_type_0 = const()[name = tensor("hidden_states_331_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_331_pad_0 = const()[name = tensor("hidden_states_331_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor down_blocks_2_resnets_1_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(349616704))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(360675968))), name = tensor("down_blocks_2_resnets_1_conv2_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor down_blocks_2_resnets_1_conv2_bias_to_fp16 = const()[name = tensor("down_blocks_2_resnets_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(360676160)))]; + tensor hidden_states_331_cast_fp16 = conv(bias = down_blocks_2_resnets_1_conv2_bias_to_fp16, dilations = var_2386, groups = var_1048, pad = hidden_states_331_pad_0, pad_type = hidden_states_331_pad_type_0, strides = var_2384, weight = down_blocks_2_resnets_1_conv2_weight_to_fp16_palettized, x = input_427_cast_fp16)[name = tensor("hidden_states_331_cast_fp16")]; + tensor var_2389_cast_fp16 = add(x = input_415_cast_fp16, y = hidden_states_331_cast_fp16)[name = tensor("op_2389_cast_fp16")]; + tensor reshape_60_shape_0 = const()[name = tensor("reshape_60_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_60_cast_fp16 = reshape(shape = reshape_60_shape_0, x = var_2389_cast_fp16)[name = tensor("reshape_60_cast_fp16")]; + tensor reduce_mean_45_axes_0 = const()[name = tensor("reduce_mean_45_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_45_keep_dims_0 = const()[name = tensor("reduce_mean_45_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_45_cast_fp16 = reduce_mean(axes = reduce_mean_45_axes_0, keep_dims = reduce_mean_45_keep_dims_0, x = reshape_60_cast_fp16)[name = tensor("reduce_mean_45_cast_fp16")]; + tensor sub_30_cast_fp16 = sub(x = reshape_60_cast_fp16, y = reduce_mean_45_cast_fp16)[name = tensor("sub_30_cast_fp16")]; + tensor square_15_cast_fp16 = square(x = sub_30_cast_fp16)[name = tensor("square_15_cast_fp16")]; + tensor reduce_mean_47_axes_0 = const()[name = tensor("reduce_mean_47_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_47_keep_dims_0 = const()[name = tensor("reduce_mean_47_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_47_cast_fp16 = reduce_mean(axes = reduce_mean_47_axes_0, keep_dims = reduce_mean_47_keep_dims_0, x = square_15_cast_fp16)[name = tensor("reduce_mean_47_cast_fp16")]; + tensor add_30_y_0_to_fp16 = const()[name = tensor("add_30_y_0_to_fp16"), val = tensor(0x1.1p-20)]; + tensor add_30_cast_fp16 = add(x = reduce_mean_47_cast_fp16, y = add_30_y_0_to_fp16)[name = tensor("add_30_cast_fp16")]; + tensor sqrt_15_cast_fp16 = sqrt(x = add_30_cast_fp16)[name = tensor("sqrt_15_cast_fp16")]; + tensor real_div_15_cast_fp16 = real_div(x = sub_30_cast_fp16, y = sqrt_15_cast_fp16)[name = tensor("real_div_15_cast_fp16")]; + tensor reshape_61_shape_0 = const()[name = tensor("reshape_61_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_61_cast_fp16 = reshape(shape = reshape_61_shape_0, x = real_div_15_cast_fp16)[name = tensor("reshape_61_cast_fp16")]; + tensor add_31_gamma_0_to_fp16 = const()[name = tensor("add_31_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(360678784)))]; + tensor add_31_beta_0_to_fp16 = const()[name = tensor("add_31_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(360681408)))]; + tensor add_31_epsilon_0_to_fp16 = const()[name = tensor("add_31_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_31_cast_fp16 = batch_norm(beta = add_31_beta_0_to_fp16, epsilon = add_31_epsilon_0_to_fp16, gamma = add_31_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_61_cast_fp16)[name = tensor("add_31_cast_fp16")]; + tensor var_2427 = const()[name = tensor("op_2427"), val = tensor([0, 2, 3, 1])]; + tensor var_2431 = const()[name = tensor("op_2431"), val = tensor([1, 1024, 1280])]; + tensor transpose_571 = transpose(perm = var_2427, x = add_31_cast_fp16)[name = tensor("transpose_571")]; + tensor input_431_cast_fp16 = reshape(shape = var_2431, x = transpose_571)[name = tensor("input_431_cast_fp16")]; + tensor down_blocks_2_attentions_1_proj_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(360684032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(361912896))), name = tensor("down_blocks_2_attentions_1_proj_in_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_proj_in_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_proj_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(361913088)))]; + tensor linear_156_cast_fp16 = linear(bias = down_blocks_2_attentions_1_proj_in_bias_to_fp16, weight = down_blocks_2_attentions_1_proj_in_weight_to_fp16_palettized, x = input_431_cast_fp16)[name = tensor("linear_156_cast_fp16")]; + tensor hidden_states_339_axes_0 = const()[name = tensor("hidden_states_339_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(361915712)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(361918336)))]; + tensor hidden_states_339_cast_fp16 = layer_norm(axes = hidden_states_339_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_0_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_0_norm1_weight_to_fp16, x = linear_156_cast_fp16)[name = tensor("hidden_states_339_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(361920960))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(363149824))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_157_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_339_cast_fp16)[name = tensor("linear_157_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(363150016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(364378880))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_158_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_339_cast_fp16)[name = tensor("linear_158_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(364379072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(365607936))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_159_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_339_cast_fp16)[name = tensor("linear_159_cast_fp16")]; + tensor var_2464 = const()[name = tensor("op_2464"), val = tensor([1, -1, 20, 64])]; + tensor var_2465_cast_fp16 = reshape(shape = var_2464, x = linear_157_cast_fp16)[name = tensor("op_2465_cast_fp16")]; + tensor var_2467 = const()[name = tensor("op_2467"), val = tensor([1, -1, 20, 64])]; + tensor var_2468_cast_fp16 = reshape(shape = var_2467, x = linear_158_cast_fp16)[name = tensor("op_2468_cast_fp16")]; + tensor var_2470 = const()[name = tensor("op_2470"), val = tensor([1, -1, 20, 64])]; + tensor var_2471_cast_fp16 = reshape(shape = var_2470, x = linear_159_cast_fp16)[name = tensor("op_2471_cast_fp16")]; + tensor value_115_perm_0 = const()[name = tensor("value_115_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_44_y_0_to_fp16 = const()[name = tensor("mul_44_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_44_cast_fp16 = mul(x = var_2465_cast_fp16, y = mul_44_y_0_to_fp16)[name = tensor("mul_44_cast_fp16")]; + tensor matmul_28_transpose_y_0 = const()[name = tensor("matmul_28_transpose_y_0"), val = tensor(true)]; + tensor matmul_28_transpose_x_0 = const()[name = tensor("matmul_28_transpose_x_0"), val = tensor(false)]; + tensor transpose_328_perm_0 = const()[name = tensor("transpose_328_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_329_perm_0 = const()[name = tensor("transpose_329_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_569 = transpose(perm = transpose_329_perm_0, x = var_2468_cast_fp16)[name = tensor("transpose_569")]; + tensor transpose_570 = transpose(perm = transpose_328_perm_0, x = mul_44_cast_fp16)[name = tensor("transpose_570")]; + tensor matmul_28_cast_fp16 = matmul(transpose_x = matmul_28_transpose_x_0, transpose_y = matmul_28_transpose_y_0, x = transpose_570, y = transpose_569)[name = tensor("matmul_28_cast_fp16")]; + tensor softmax_28_axis_0 = const()[name = tensor("softmax_28_axis_0"), val = tensor(-1)]; + tensor softmax_28_cast_fp16 = softmax(axis = softmax_28_axis_0, x = matmul_28_cast_fp16)[name = tensor("softmax_28_cast_fp16")]; + tensor hidden_states_341_transpose_x_0 = const()[name = tensor("hidden_states_341_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_341_transpose_y_0 = const()[name = tensor("hidden_states_341_transpose_y_0"), val = tensor(false)]; + tensor transpose_568 = transpose(perm = value_115_perm_0, x = var_2471_cast_fp16)[name = tensor("transpose_568")]; + tensor hidden_states_341_cast_fp16 = matmul(transpose_x = hidden_states_341_transpose_x_0, transpose_y = hidden_states_341_transpose_y_0, x = softmax_28_cast_fp16, y = transpose_568)[name = tensor("hidden_states_341_cast_fp16")]; + tensor var_2474_perm_0 = const()[name = tensor("op_2474_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2478 = const()[name = tensor("op_2478"), val = tensor([1, -1, 1280])]; + tensor transpose_567 = transpose(perm = var_2474_perm_0, x = hidden_states_341_cast_fp16)[name = tensor("transpose_567")]; + tensor hidden_states_343_cast_fp16 = reshape(shape = var_2478, x = transpose_567)[name = tensor("hidden_states_343_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(365608128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(366836992))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(366837184)))]; + tensor linear_160_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_343_cast_fp16)[name = tensor("linear_160_cast_fp16")]; + tensor input_437_cast_fp16 = add(x = linear_160_cast_fp16, y = linear_156_cast_fp16)[name = tensor("input_437_cast_fp16")]; + tensor input_439_axes_0 = const()[name = tensor("input_439_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(366839808)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(366842432)))]; + tensor input_439_cast_fp16 = layer_norm(axes = input_439_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_0_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_0_norm2_weight_to_fp16, x = input_437_cast_fp16)[name = tensor("input_439_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(366845056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(368073920))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_161_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized, x = input_439_cast_fp16)[name = tensor("linear_161_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(368074112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(370040256))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_162_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_162_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(370040448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(372006592))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_163_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_163_cast_fp16")]; + tensor var_2510 = const()[name = tensor("op_2510"), val = tensor([1, -1, 20, 64])]; + tensor var_2511_cast_fp16 = reshape(shape = var_2510, x = linear_161_cast_fp16)[name = tensor("op_2511_cast_fp16")]; + tensor var_2513 = const()[name = tensor("op_2513"), val = tensor([1, -1, 20, 64])]; + tensor var_2514_cast_fp16 = reshape(shape = var_2513, x = linear_162_cast_fp16)[name = tensor("op_2514_cast_fp16")]; + tensor var_2516 = const()[name = tensor("op_2516"), val = tensor([1, -1, 20, 64])]; + tensor var_2517_cast_fp16 = reshape(shape = var_2516, x = linear_163_cast_fp16)[name = tensor("op_2517_cast_fp16")]; + tensor value_119_perm_0 = const()[name = tensor("value_119_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_45_y_0_to_fp16 = const()[name = tensor("mul_45_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_45_cast_fp16 = mul(x = var_2511_cast_fp16, y = mul_45_y_0_to_fp16)[name = tensor("mul_45_cast_fp16")]; + tensor matmul_29_transpose_y_0 = const()[name = tensor("matmul_29_transpose_y_0"), val = tensor(true)]; + tensor matmul_29_transpose_x_0 = const()[name = tensor("matmul_29_transpose_x_0"), val = tensor(false)]; + tensor transpose_330_perm_0 = const()[name = tensor("transpose_330_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_331_perm_0 = const()[name = tensor("transpose_331_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_565 = transpose(perm = transpose_331_perm_0, x = var_2514_cast_fp16)[name = tensor("transpose_565")]; + tensor transpose_566 = transpose(perm = transpose_330_perm_0, x = mul_45_cast_fp16)[name = tensor("transpose_566")]; + tensor matmul_29_cast_fp16 = matmul(transpose_x = matmul_29_transpose_x_0, transpose_y = matmul_29_transpose_y_0, x = transpose_566, y = transpose_565)[name = tensor("matmul_29_cast_fp16")]; + tensor softmax_29_axis_0 = const()[name = tensor("softmax_29_axis_0"), val = tensor(-1)]; + tensor softmax_29_cast_fp16 = softmax(axis = softmax_29_axis_0, x = matmul_29_cast_fp16)[name = tensor("softmax_29_cast_fp16")]; + tensor hidden_states_347_transpose_x_0 = const()[name = tensor("hidden_states_347_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_347_transpose_y_0 = const()[name = tensor("hidden_states_347_transpose_y_0"), val = tensor(false)]; + tensor transpose_564 = transpose(perm = value_119_perm_0, x = var_2517_cast_fp16)[name = tensor("transpose_564")]; + tensor hidden_states_347_cast_fp16 = matmul(transpose_x = hidden_states_347_transpose_x_0, transpose_y = hidden_states_347_transpose_y_0, x = softmax_29_cast_fp16, y = transpose_564)[name = tensor("hidden_states_347_cast_fp16")]; + tensor var_2520_perm_0 = const()[name = tensor("op_2520_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2524 = const()[name = tensor("op_2524"), val = tensor([1, -1, 1280])]; + tensor transpose_563 = transpose(perm = var_2520_perm_0, x = hidden_states_347_cast_fp16)[name = tensor("transpose_563")]; + tensor hidden_states_349_cast_fp16 = reshape(shape = var_2524, x = transpose_563)[name = tensor("hidden_states_349_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(372006784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(373235648))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(373235840)))]; + tensor linear_164_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_349_cast_fp16)[name = tensor("linear_164_cast_fp16")]; + tensor input_445_cast_fp16 = add(x = linear_164_cast_fp16, y = input_437_cast_fp16)[name = tensor("input_445_cast_fp16")]; + tensor input_447_axes_0 = const()[name = tensor("input_447_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(373238464)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(373241088)))]; + tensor input_447_cast_fp16 = layer_norm(axes = input_447_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_0_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_0_norm3_weight_to_fp16, x = input_445_cast_fp16)[name = tensor("input_447_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(373243712))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(383074176))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(383074368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(383082112))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_165_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized, x = input_447_cast_fp16)[name = tensor("linear_165_cast_fp16")]; + tensor var_2546_split_sizes_0 = const()[name = tensor("op_2546_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2546_axis_0 = const()[name = tensor("op_2546_axis_0"), val = tensor(-1)]; + tensor var_2546_cast_fp16_0, tensor var_2546_cast_fp16_1 = split(axis = var_2546_axis_0, split_sizes = var_2546_split_sizes_0, x = linear_165_cast_fp16)[name = tensor("op_2546_cast_fp16")]; + tensor var_2548_mode_0 = const()[name = tensor("op_2548_mode_0"), val = tensor("EXACT")]; + tensor var_2548_cast_fp16 = gelu(mode = var_2548_mode_0, x = var_2546_cast_fp16_1)[name = tensor("op_2548_cast_fp16")]; + tensor input_449_cast_fp16 = mul(x = var_2546_cast_fp16_0, y = var_2548_cast_fp16)[name = tensor("input_449_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(383082304))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(387997568))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(387997760)))]; + tensor linear_166_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized, x = input_449_cast_fp16)[name = tensor("linear_166_cast_fp16")]; + tensor hidden_states_357_cast_fp16 = add(x = linear_166_cast_fp16, y = input_445_cast_fp16)[name = tensor("hidden_states_357_cast_fp16")]; + tensor hidden_states_359_axes_0 = const()[name = tensor("hidden_states_359_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(388000384)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(388003008)))]; + tensor hidden_states_359_cast_fp16 = layer_norm(axes = hidden_states_359_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_1_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_1_norm1_weight_to_fp16, x = hidden_states_357_cast_fp16)[name = tensor("hidden_states_359_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(388005632))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(389234496))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_167_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_359_cast_fp16)[name = tensor("linear_167_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(389234688))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(390463552))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_168_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_359_cast_fp16)[name = tensor("linear_168_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(390463744))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(391692608))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_169_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_359_cast_fp16)[name = tensor("linear_169_cast_fp16")]; + tensor var_2583 = const()[name = tensor("op_2583"), val = tensor([1, -1, 20, 64])]; + tensor var_2584_cast_fp16 = reshape(shape = var_2583, x = linear_167_cast_fp16)[name = tensor("op_2584_cast_fp16")]; + tensor var_2586 = const()[name = tensor("op_2586"), val = tensor([1, -1, 20, 64])]; + tensor var_2587_cast_fp16 = reshape(shape = var_2586, x = linear_168_cast_fp16)[name = tensor("op_2587_cast_fp16")]; + tensor var_2589 = const()[name = tensor("op_2589"), val = tensor([1, -1, 20, 64])]; + tensor var_2590_cast_fp16 = reshape(shape = var_2589, x = linear_169_cast_fp16)[name = tensor("op_2590_cast_fp16")]; + tensor value_123_perm_0 = const()[name = tensor("value_123_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_46_y_0_to_fp16 = const()[name = tensor("mul_46_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_46_cast_fp16 = mul(x = var_2584_cast_fp16, y = mul_46_y_0_to_fp16)[name = tensor("mul_46_cast_fp16")]; + tensor matmul_30_transpose_y_0 = const()[name = tensor("matmul_30_transpose_y_0"), val = tensor(true)]; + tensor matmul_30_transpose_x_0 = const()[name = tensor("matmul_30_transpose_x_0"), val = tensor(false)]; + tensor transpose_332_perm_0 = const()[name = tensor("transpose_332_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_333_perm_0 = const()[name = tensor("transpose_333_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_561 = transpose(perm = transpose_333_perm_0, x = var_2587_cast_fp16)[name = tensor("transpose_561")]; + tensor transpose_562 = transpose(perm = transpose_332_perm_0, x = mul_46_cast_fp16)[name = tensor("transpose_562")]; + tensor matmul_30_cast_fp16 = matmul(transpose_x = matmul_30_transpose_x_0, transpose_y = matmul_30_transpose_y_0, x = transpose_562, y = transpose_561)[name = tensor("matmul_30_cast_fp16")]; + tensor softmax_30_axis_0 = const()[name = tensor("softmax_30_axis_0"), val = tensor(-1)]; + tensor softmax_30_cast_fp16 = softmax(axis = softmax_30_axis_0, x = matmul_30_cast_fp16)[name = tensor("softmax_30_cast_fp16")]; + tensor hidden_states_361_transpose_x_0 = const()[name = tensor("hidden_states_361_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_361_transpose_y_0 = const()[name = tensor("hidden_states_361_transpose_y_0"), val = tensor(false)]; + tensor transpose_560 = transpose(perm = value_123_perm_0, x = var_2590_cast_fp16)[name = tensor("transpose_560")]; + tensor hidden_states_361_cast_fp16 = matmul(transpose_x = hidden_states_361_transpose_x_0, transpose_y = hidden_states_361_transpose_y_0, x = softmax_30_cast_fp16, y = transpose_560)[name = tensor("hidden_states_361_cast_fp16")]; + tensor var_2593_perm_0 = const()[name = tensor("op_2593_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2597 = const()[name = tensor("op_2597"), val = tensor([1, -1, 1280])]; + tensor transpose_559 = transpose(perm = var_2593_perm_0, x = hidden_states_361_cast_fp16)[name = tensor("transpose_559")]; + tensor hidden_states_363_cast_fp16 = reshape(shape = var_2597, x = transpose_559)[name = tensor("hidden_states_363_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(391692800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(392921664))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(392921856)))]; + tensor linear_170_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_363_cast_fp16)[name = tensor("linear_170_cast_fp16")]; + tensor input_457_cast_fp16 = add(x = linear_170_cast_fp16, y = hidden_states_357_cast_fp16)[name = tensor("input_457_cast_fp16")]; + tensor input_459_axes_0 = const()[name = tensor("input_459_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(392924480)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(392927104)))]; + tensor input_459_cast_fp16 = layer_norm(axes = input_459_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_1_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_1_norm2_weight_to_fp16, x = input_457_cast_fp16)[name = tensor("input_459_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(392929728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(394158592))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_171_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized, x = input_459_cast_fp16)[name = tensor("linear_171_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(394158784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(396124928))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_172_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_172_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(396125120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(398091264))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_173_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_173_cast_fp16")]; + tensor var_2629 = const()[name = tensor("op_2629"), val = tensor([1, -1, 20, 64])]; + tensor var_2630_cast_fp16 = reshape(shape = var_2629, x = linear_171_cast_fp16)[name = tensor("op_2630_cast_fp16")]; + tensor var_2632 = const()[name = tensor("op_2632"), val = tensor([1, -1, 20, 64])]; + tensor var_2633_cast_fp16 = reshape(shape = var_2632, x = linear_172_cast_fp16)[name = tensor("op_2633_cast_fp16")]; + tensor var_2635 = const()[name = tensor("op_2635"), val = tensor([1, -1, 20, 64])]; + tensor var_2636_cast_fp16 = reshape(shape = var_2635, x = linear_173_cast_fp16)[name = tensor("op_2636_cast_fp16")]; + tensor value_127_perm_0 = const()[name = tensor("value_127_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_47_y_0_to_fp16 = const()[name = tensor("mul_47_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_47_cast_fp16 = mul(x = var_2630_cast_fp16, y = mul_47_y_0_to_fp16)[name = tensor("mul_47_cast_fp16")]; + tensor matmul_31_transpose_y_0 = const()[name = tensor("matmul_31_transpose_y_0"), val = tensor(true)]; + tensor matmul_31_transpose_x_0 = const()[name = tensor("matmul_31_transpose_x_0"), val = tensor(false)]; + tensor transpose_334_perm_0 = const()[name = tensor("transpose_334_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_335_perm_0 = const()[name = tensor("transpose_335_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_557 = transpose(perm = transpose_335_perm_0, x = var_2633_cast_fp16)[name = tensor("transpose_557")]; + tensor transpose_558 = transpose(perm = transpose_334_perm_0, x = mul_47_cast_fp16)[name = tensor("transpose_558")]; + tensor matmul_31_cast_fp16 = matmul(transpose_x = matmul_31_transpose_x_0, transpose_y = matmul_31_transpose_y_0, x = transpose_558, y = transpose_557)[name = tensor("matmul_31_cast_fp16")]; + tensor softmax_31_axis_0 = const()[name = tensor("softmax_31_axis_0"), val = tensor(-1)]; + tensor softmax_31_cast_fp16 = softmax(axis = softmax_31_axis_0, x = matmul_31_cast_fp16)[name = tensor("softmax_31_cast_fp16")]; + tensor hidden_states_367_transpose_x_0 = const()[name = tensor("hidden_states_367_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_367_transpose_y_0 = const()[name = tensor("hidden_states_367_transpose_y_0"), val = tensor(false)]; + tensor transpose_556 = transpose(perm = value_127_perm_0, x = var_2636_cast_fp16)[name = tensor("transpose_556")]; + tensor hidden_states_367_cast_fp16 = matmul(transpose_x = hidden_states_367_transpose_x_0, transpose_y = hidden_states_367_transpose_y_0, x = softmax_31_cast_fp16, y = transpose_556)[name = tensor("hidden_states_367_cast_fp16")]; + tensor var_2639_perm_0 = const()[name = tensor("op_2639_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2643 = const()[name = tensor("op_2643"), val = tensor([1, -1, 1280])]; + tensor transpose_555 = transpose(perm = var_2639_perm_0, x = hidden_states_367_cast_fp16)[name = tensor("transpose_555")]; + tensor hidden_states_369_cast_fp16 = reshape(shape = var_2643, x = transpose_555)[name = tensor("hidden_states_369_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(398091456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(399320320))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(399320512)))]; + tensor linear_174_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_369_cast_fp16)[name = tensor("linear_174_cast_fp16")]; + tensor input_465_cast_fp16 = add(x = linear_174_cast_fp16, y = input_457_cast_fp16)[name = tensor("input_465_cast_fp16")]; + tensor input_467_axes_0 = const()[name = tensor("input_467_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(399323136)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(399325760)))]; + tensor input_467_cast_fp16 = layer_norm(axes = input_467_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_1_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_1_norm3_weight_to_fp16, x = input_465_cast_fp16)[name = tensor("input_467_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(399328384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(409158848))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(409159040))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(409166784))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_175_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized, x = input_467_cast_fp16)[name = tensor("linear_175_cast_fp16")]; + tensor var_2665_split_sizes_0 = const()[name = tensor("op_2665_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2665_axis_0 = const()[name = tensor("op_2665_axis_0"), val = tensor(-1)]; + tensor var_2665_cast_fp16_0, tensor var_2665_cast_fp16_1 = split(axis = var_2665_axis_0, split_sizes = var_2665_split_sizes_0, x = linear_175_cast_fp16)[name = tensor("op_2665_cast_fp16")]; + tensor var_2667_mode_0 = const()[name = tensor("op_2667_mode_0"), val = tensor("EXACT")]; + tensor var_2667_cast_fp16 = gelu(mode = var_2667_mode_0, x = var_2665_cast_fp16_1)[name = tensor("op_2667_cast_fp16")]; + tensor input_469_cast_fp16 = mul(x = var_2665_cast_fp16_0, y = var_2667_cast_fp16)[name = tensor("input_469_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(409166976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(414082240))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(414082432)))]; + tensor linear_176_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized, x = input_469_cast_fp16)[name = tensor("linear_176_cast_fp16")]; + tensor hidden_states_377_cast_fp16 = add(x = linear_176_cast_fp16, y = input_465_cast_fp16)[name = tensor("hidden_states_377_cast_fp16")]; + tensor hidden_states_379_axes_0 = const()[name = tensor("hidden_states_379_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(414085056)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(414087680)))]; + tensor hidden_states_379_cast_fp16 = layer_norm(axes = hidden_states_379_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_2_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_2_norm1_weight_to_fp16, x = hidden_states_377_cast_fp16)[name = tensor("hidden_states_379_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(414090304))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(415319168))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_177_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_379_cast_fp16)[name = tensor("linear_177_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(415319360))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(416548224))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_178_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_379_cast_fp16)[name = tensor("linear_178_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(416548416))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(417777280))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_179_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_379_cast_fp16)[name = tensor("linear_179_cast_fp16")]; + tensor var_2702 = const()[name = tensor("op_2702"), val = tensor([1, -1, 20, 64])]; + tensor var_2703_cast_fp16 = reshape(shape = var_2702, x = linear_177_cast_fp16)[name = tensor("op_2703_cast_fp16")]; + tensor var_2705 = const()[name = tensor("op_2705"), val = tensor([1, -1, 20, 64])]; + tensor var_2706_cast_fp16 = reshape(shape = var_2705, x = linear_178_cast_fp16)[name = tensor("op_2706_cast_fp16")]; + tensor var_2708 = const()[name = tensor("op_2708"), val = tensor([1, -1, 20, 64])]; + tensor var_2709_cast_fp16 = reshape(shape = var_2708, x = linear_179_cast_fp16)[name = tensor("op_2709_cast_fp16")]; + tensor value_131_perm_0 = const()[name = tensor("value_131_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_48_y_0_to_fp16 = const()[name = tensor("mul_48_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_48_cast_fp16 = mul(x = var_2703_cast_fp16, y = mul_48_y_0_to_fp16)[name = tensor("mul_48_cast_fp16")]; + tensor matmul_32_transpose_y_0 = const()[name = tensor("matmul_32_transpose_y_0"), val = tensor(true)]; + tensor matmul_32_transpose_x_0 = const()[name = tensor("matmul_32_transpose_x_0"), val = tensor(false)]; + tensor transpose_336_perm_0 = const()[name = tensor("transpose_336_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_337_perm_0 = const()[name = tensor("transpose_337_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_553 = transpose(perm = transpose_337_perm_0, x = var_2706_cast_fp16)[name = tensor("transpose_553")]; + tensor transpose_554 = transpose(perm = transpose_336_perm_0, x = mul_48_cast_fp16)[name = tensor("transpose_554")]; + tensor matmul_32_cast_fp16 = matmul(transpose_x = matmul_32_transpose_x_0, transpose_y = matmul_32_transpose_y_0, x = transpose_554, y = transpose_553)[name = tensor("matmul_32_cast_fp16")]; + tensor softmax_32_axis_0 = const()[name = tensor("softmax_32_axis_0"), val = tensor(-1)]; + tensor softmax_32_cast_fp16 = softmax(axis = softmax_32_axis_0, x = matmul_32_cast_fp16)[name = tensor("softmax_32_cast_fp16")]; + tensor hidden_states_381_transpose_x_0 = const()[name = tensor("hidden_states_381_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_381_transpose_y_0 = const()[name = tensor("hidden_states_381_transpose_y_0"), val = tensor(false)]; + tensor transpose_552 = transpose(perm = value_131_perm_0, x = var_2709_cast_fp16)[name = tensor("transpose_552")]; + tensor hidden_states_381_cast_fp16 = matmul(transpose_x = hidden_states_381_transpose_x_0, transpose_y = hidden_states_381_transpose_y_0, x = softmax_32_cast_fp16, y = transpose_552)[name = tensor("hidden_states_381_cast_fp16")]; + tensor var_2712_perm_0 = const()[name = tensor("op_2712_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2716 = const()[name = tensor("op_2716"), val = tensor([1, -1, 1280])]; + tensor transpose_551 = transpose(perm = var_2712_perm_0, x = hidden_states_381_cast_fp16)[name = tensor("transpose_551")]; + tensor hidden_states_383_cast_fp16 = reshape(shape = var_2716, x = transpose_551)[name = tensor("hidden_states_383_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(417777472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(419006336))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(419006528)))]; + tensor linear_180_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_383_cast_fp16)[name = tensor("linear_180_cast_fp16")]; + tensor input_477_cast_fp16 = add(x = linear_180_cast_fp16, y = hidden_states_377_cast_fp16)[name = tensor("input_477_cast_fp16")]; + tensor input_479_axes_0 = const()[name = tensor("input_479_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(419009152)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(419011776)))]; + tensor input_479_cast_fp16 = layer_norm(axes = input_479_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_2_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_2_norm2_weight_to_fp16, x = input_477_cast_fp16)[name = tensor("input_479_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(419014400))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(420243264))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_181_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized, x = input_479_cast_fp16)[name = tensor("linear_181_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(420243456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(422209600))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_182_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_182_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(422209792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(424175936))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_183_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_183_cast_fp16")]; + tensor var_2748 = const()[name = tensor("op_2748"), val = tensor([1, -1, 20, 64])]; + tensor var_2749_cast_fp16 = reshape(shape = var_2748, x = linear_181_cast_fp16)[name = tensor("op_2749_cast_fp16")]; + tensor var_2751 = const()[name = tensor("op_2751"), val = tensor([1, -1, 20, 64])]; + tensor var_2752_cast_fp16 = reshape(shape = var_2751, x = linear_182_cast_fp16)[name = tensor("op_2752_cast_fp16")]; + tensor var_2754 = const()[name = tensor("op_2754"), val = tensor([1, -1, 20, 64])]; + tensor var_2755_cast_fp16 = reshape(shape = var_2754, x = linear_183_cast_fp16)[name = tensor("op_2755_cast_fp16")]; + tensor value_135_perm_0 = const()[name = tensor("value_135_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_49_y_0_to_fp16 = const()[name = tensor("mul_49_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_49_cast_fp16 = mul(x = var_2749_cast_fp16, y = mul_49_y_0_to_fp16)[name = tensor("mul_49_cast_fp16")]; + tensor matmul_33_transpose_y_0 = const()[name = tensor("matmul_33_transpose_y_0"), val = tensor(true)]; + tensor matmul_33_transpose_x_0 = const()[name = tensor("matmul_33_transpose_x_0"), val = tensor(false)]; + tensor transpose_338_perm_0 = const()[name = tensor("transpose_338_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_339_perm_0 = const()[name = tensor("transpose_339_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_549 = transpose(perm = transpose_339_perm_0, x = var_2752_cast_fp16)[name = tensor("transpose_549")]; + tensor transpose_550 = transpose(perm = transpose_338_perm_0, x = mul_49_cast_fp16)[name = tensor("transpose_550")]; + tensor matmul_33_cast_fp16 = matmul(transpose_x = matmul_33_transpose_x_0, transpose_y = matmul_33_transpose_y_0, x = transpose_550, y = transpose_549)[name = tensor("matmul_33_cast_fp16")]; + tensor softmax_33_axis_0 = const()[name = tensor("softmax_33_axis_0"), val = tensor(-1)]; + tensor softmax_33_cast_fp16 = softmax(axis = softmax_33_axis_0, x = matmul_33_cast_fp16)[name = tensor("softmax_33_cast_fp16")]; + tensor hidden_states_387_transpose_x_0 = const()[name = tensor("hidden_states_387_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_387_transpose_y_0 = const()[name = tensor("hidden_states_387_transpose_y_0"), val = tensor(false)]; + tensor transpose_548 = transpose(perm = value_135_perm_0, x = var_2755_cast_fp16)[name = tensor("transpose_548")]; + tensor hidden_states_387_cast_fp16 = matmul(transpose_x = hidden_states_387_transpose_x_0, transpose_y = hidden_states_387_transpose_y_0, x = softmax_33_cast_fp16, y = transpose_548)[name = tensor("hidden_states_387_cast_fp16")]; + tensor var_2758_perm_0 = const()[name = tensor("op_2758_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2762 = const()[name = tensor("op_2762"), val = tensor([1, -1, 1280])]; + tensor transpose_547 = transpose(perm = var_2758_perm_0, x = hidden_states_387_cast_fp16)[name = tensor("transpose_547")]; + tensor hidden_states_389_cast_fp16 = reshape(shape = var_2762, x = transpose_547)[name = tensor("hidden_states_389_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(424176128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(425404992))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(425405184)))]; + tensor linear_184_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_389_cast_fp16)[name = tensor("linear_184_cast_fp16")]; + tensor input_485_cast_fp16 = add(x = linear_184_cast_fp16, y = input_477_cast_fp16)[name = tensor("input_485_cast_fp16")]; + tensor input_487_axes_0 = const()[name = tensor("input_487_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(425407808)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(425410432)))]; + tensor input_487_cast_fp16 = layer_norm(axes = input_487_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_2_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_2_norm3_weight_to_fp16, x = input_485_cast_fp16)[name = tensor("input_487_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(425413056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(435243520))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(435243712))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(435251456))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_185_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized, x = input_487_cast_fp16)[name = tensor("linear_185_cast_fp16")]; + tensor var_2784_split_sizes_0 = const()[name = tensor("op_2784_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2784_axis_0 = const()[name = tensor("op_2784_axis_0"), val = tensor(-1)]; + tensor var_2784_cast_fp16_0, tensor var_2784_cast_fp16_1 = split(axis = var_2784_axis_0, split_sizes = var_2784_split_sizes_0, x = linear_185_cast_fp16)[name = tensor("op_2784_cast_fp16")]; + tensor var_2786_mode_0 = const()[name = tensor("op_2786_mode_0"), val = tensor("EXACT")]; + tensor var_2786_cast_fp16 = gelu(mode = var_2786_mode_0, x = var_2784_cast_fp16_1)[name = tensor("op_2786_cast_fp16")]; + tensor input_489_cast_fp16 = mul(x = var_2784_cast_fp16_0, y = var_2786_cast_fp16)[name = tensor("input_489_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(435251648))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(440166912))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(440167104)))]; + tensor linear_186_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized, x = input_489_cast_fp16)[name = tensor("linear_186_cast_fp16")]; + tensor hidden_states_397_cast_fp16 = add(x = linear_186_cast_fp16, y = input_485_cast_fp16)[name = tensor("hidden_states_397_cast_fp16")]; + tensor hidden_states_399_axes_0 = const()[name = tensor("hidden_states_399_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(440169728)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(440172352)))]; + tensor hidden_states_399_cast_fp16 = layer_norm(axes = hidden_states_399_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_3_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_3_norm1_weight_to_fp16, x = hidden_states_397_cast_fp16)[name = tensor("hidden_states_399_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(440174976))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(441403840))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_187_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_399_cast_fp16)[name = tensor("linear_187_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(441404032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(442632896))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_188_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_399_cast_fp16)[name = tensor("linear_188_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(442633088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(443861952))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_189_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_399_cast_fp16)[name = tensor("linear_189_cast_fp16")]; + tensor var_2821 = const()[name = tensor("op_2821"), val = tensor([1, -1, 20, 64])]; + tensor var_2822_cast_fp16 = reshape(shape = var_2821, x = linear_187_cast_fp16)[name = tensor("op_2822_cast_fp16")]; + tensor var_2824 = const()[name = tensor("op_2824"), val = tensor([1, -1, 20, 64])]; + tensor var_2825_cast_fp16 = reshape(shape = var_2824, x = linear_188_cast_fp16)[name = tensor("op_2825_cast_fp16")]; + tensor var_2827 = const()[name = tensor("op_2827"), val = tensor([1, -1, 20, 64])]; + tensor var_2828_cast_fp16 = reshape(shape = var_2827, x = linear_189_cast_fp16)[name = tensor("op_2828_cast_fp16")]; + tensor value_139_perm_0 = const()[name = tensor("value_139_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_50_y_0_to_fp16 = const()[name = tensor("mul_50_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_50_cast_fp16 = mul(x = var_2822_cast_fp16, y = mul_50_y_0_to_fp16)[name = tensor("mul_50_cast_fp16")]; + tensor matmul_34_transpose_y_0 = const()[name = tensor("matmul_34_transpose_y_0"), val = tensor(true)]; + tensor matmul_34_transpose_x_0 = const()[name = tensor("matmul_34_transpose_x_0"), val = tensor(false)]; + tensor transpose_340_perm_0 = const()[name = tensor("transpose_340_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_341_perm_0 = const()[name = tensor("transpose_341_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_545 = transpose(perm = transpose_341_perm_0, x = var_2825_cast_fp16)[name = tensor("transpose_545")]; + tensor transpose_546 = transpose(perm = transpose_340_perm_0, x = mul_50_cast_fp16)[name = tensor("transpose_546")]; + tensor matmul_34_cast_fp16 = matmul(transpose_x = matmul_34_transpose_x_0, transpose_y = matmul_34_transpose_y_0, x = transpose_546, y = transpose_545)[name = tensor("matmul_34_cast_fp16")]; + tensor softmax_34_axis_0 = const()[name = tensor("softmax_34_axis_0"), val = tensor(-1)]; + tensor softmax_34_cast_fp16 = softmax(axis = softmax_34_axis_0, x = matmul_34_cast_fp16)[name = tensor("softmax_34_cast_fp16")]; + tensor hidden_states_401_transpose_x_0 = const()[name = tensor("hidden_states_401_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_401_transpose_y_0 = const()[name = tensor("hidden_states_401_transpose_y_0"), val = tensor(false)]; + tensor transpose_544 = transpose(perm = value_139_perm_0, x = var_2828_cast_fp16)[name = tensor("transpose_544")]; + tensor hidden_states_401_cast_fp16 = matmul(transpose_x = hidden_states_401_transpose_x_0, transpose_y = hidden_states_401_transpose_y_0, x = softmax_34_cast_fp16, y = transpose_544)[name = tensor("hidden_states_401_cast_fp16")]; + tensor var_2831_perm_0 = const()[name = tensor("op_2831_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2835 = const()[name = tensor("op_2835"), val = tensor([1, -1, 1280])]; + tensor transpose_543 = transpose(perm = var_2831_perm_0, x = hidden_states_401_cast_fp16)[name = tensor("transpose_543")]; + tensor hidden_states_403_cast_fp16 = reshape(shape = var_2835, x = transpose_543)[name = tensor("hidden_states_403_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(443862144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(445091008))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(445091200)))]; + tensor linear_190_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_403_cast_fp16)[name = tensor("linear_190_cast_fp16")]; + tensor input_497_cast_fp16 = add(x = linear_190_cast_fp16, y = hidden_states_397_cast_fp16)[name = tensor("input_497_cast_fp16")]; + tensor input_499_axes_0 = const()[name = tensor("input_499_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(445093824)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(445096448)))]; + tensor input_499_cast_fp16 = layer_norm(axes = input_499_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_3_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_3_norm2_weight_to_fp16, x = input_497_cast_fp16)[name = tensor("input_499_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(445099072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(446327936))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_191_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized, x = input_499_cast_fp16)[name = tensor("linear_191_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(446328128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(448294272))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_192_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_192_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(448294464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(450260608))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_193_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_193_cast_fp16")]; + tensor var_2867 = const()[name = tensor("op_2867"), val = tensor([1, -1, 20, 64])]; + tensor var_2868_cast_fp16 = reshape(shape = var_2867, x = linear_191_cast_fp16)[name = tensor("op_2868_cast_fp16")]; + tensor var_2870 = const()[name = tensor("op_2870"), val = tensor([1, -1, 20, 64])]; + tensor var_2871_cast_fp16 = reshape(shape = var_2870, x = linear_192_cast_fp16)[name = tensor("op_2871_cast_fp16")]; + tensor var_2873 = const()[name = tensor("op_2873"), val = tensor([1, -1, 20, 64])]; + tensor var_2874_cast_fp16 = reshape(shape = var_2873, x = linear_193_cast_fp16)[name = tensor("op_2874_cast_fp16")]; + tensor value_143_perm_0 = const()[name = tensor("value_143_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_51_y_0_to_fp16 = const()[name = tensor("mul_51_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_51_cast_fp16 = mul(x = var_2868_cast_fp16, y = mul_51_y_0_to_fp16)[name = tensor("mul_51_cast_fp16")]; + tensor matmul_35_transpose_y_0 = const()[name = tensor("matmul_35_transpose_y_0"), val = tensor(true)]; + tensor matmul_35_transpose_x_0 = const()[name = tensor("matmul_35_transpose_x_0"), val = tensor(false)]; + tensor transpose_342_perm_0 = const()[name = tensor("transpose_342_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_343_perm_0 = const()[name = tensor("transpose_343_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_541 = transpose(perm = transpose_343_perm_0, x = var_2871_cast_fp16)[name = tensor("transpose_541")]; + tensor transpose_542 = transpose(perm = transpose_342_perm_0, x = mul_51_cast_fp16)[name = tensor("transpose_542")]; + tensor matmul_35_cast_fp16 = matmul(transpose_x = matmul_35_transpose_x_0, transpose_y = matmul_35_transpose_y_0, x = transpose_542, y = transpose_541)[name = tensor("matmul_35_cast_fp16")]; + tensor softmax_35_axis_0 = const()[name = tensor("softmax_35_axis_0"), val = tensor(-1)]; + tensor softmax_35_cast_fp16 = softmax(axis = softmax_35_axis_0, x = matmul_35_cast_fp16)[name = tensor("softmax_35_cast_fp16")]; + tensor hidden_states_407_transpose_x_0 = const()[name = tensor("hidden_states_407_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_407_transpose_y_0 = const()[name = tensor("hidden_states_407_transpose_y_0"), val = tensor(false)]; + tensor transpose_540 = transpose(perm = value_143_perm_0, x = var_2874_cast_fp16)[name = tensor("transpose_540")]; + tensor hidden_states_407_cast_fp16 = matmul(transpose_x = hidden_states_407_transpose_x_0, transpose_y = hidden_states_407_transpose_y_0, x = softmax_35_cast_fp16, y = transpose_540)[name = tensor("hidden_states_407_cast_fp16")]; + tensor var_2877_perm_0 = const()[name = tensor("op_2877_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2881 = const()[name = tensor("op_2881"), val = tensor([1, -1, 1280])]; + tensor transpose_539 = transpose(perm = var_2877_perm_0, x = hidden_states_407_cast_fp16)[name = tensor("transpose_539")]; + tensor hidden_states_409_cast_fp16 = reshape(shape = var_2881, x = transpose_539)[name = tensor("hidden_states_409_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(450260800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(451489664))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(451489856)))]; + tensor linear_194_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_409_cast_fp16)[name = tensor("linear_194_cast_fp16")]; + tensor input_505_cast_fp16 = add(x = linear_194_cast_fp16, y = input_497_cast_fp16)[name = tensor("input_505_cast_fp16")]; + tensor input_507_axes_0 = const()[name = tensor("input_507_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(451492480)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(451495104)))]; + tensor input_507_cast_fp16 = layer_norm(axes = input_507_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_3_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_3_norm3_weight_to_fp16, x = input_505_cast_fp16)[name = tensor("input_507_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(451497728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(461328192))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(461328384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(461336128))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_195_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized, x = input_507_cast_fp16)[name = tensor("linear_195_cast_fp16")]; + tensor var_2903_split_sizes_0 = const()[name = tensor("op_2903_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_2903_axis_0 = const()[name = tensor("op_2903_axis_0"), val = tensor(-1)]; + tensor var_2903_cast_fp16_0, tensor var_2903_cast_fp16_1 = split(axis = var_2903_axis_0, split_sizes = var_2903_split_sizes_0, x = linear_195_cast_fp16)[name = tensor("op_2903_cast_fp16")]; + tensor var_2905_mode_0 = const()[name = tensor("op_2905_mode_0"), val = tensor("EXACT")]; + tensor var_2905_cast_fp16 = gelu(mode = var_2905_mode_0, x = var_2903_cast_fp16_1)[name = tensor("op_2905_cast_fp16")]; + tensor input_509_cast_fp16 = mul(x = var_2903_cast_fp16_0, y = var_2905_cast_fp16)[name = tensor("input_509_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(461336320))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(466251584))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(466251776)))]; + tensor linear_196_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized, x = input_509_cast_fp16)[name = tensor("linear_196_cast_fp16")]; + tensor hidden_states_417_cast_fp16 = add(x = linear_196_cast_fp16, y = input_505_cast_fp16)[name = tensor("hidden_states_417_cast_fp16")]; + tensor hidden_states_419_axes_0 = const()[name = tensor("hidden_states_419_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(466254400)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(466257024)))]; + tensor hidden_states_419_cast_fp16 = layer_norm(axes = hidden_states_419_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_4_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_4_norm1_weight_to_fp16, x = hidden_states_417_cast_fp16)[name = tensor("hidden_states_419_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(466259648))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(467488512))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_197_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_419_cast_fp16)[name = tensor("linear_197_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(467488704))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(468717568))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_198_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_419_cast_fp16)[name = tensor("linear_198_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(468717760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(469946624))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_199_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_419_cast_fp16)[name = tensor("linear_199_cast_fp16")]; + tensor var_2940 = const()[name = tensor("op_2940"), val = tensor([1, -1, 20, 64])]; + tensor var_2941_cast_fp16 = reshape(shape = var_2940, x = linear_197_cast_fp16)[name = tensor("op_2941_cast_fp16")]; + tensor var_2943 = const()[name = tensor("op_2943"), val = tensor([1, -1, 20, 64])]; + tensor var_2944_cast_fp16 = reshape(shape = var_2943, x = linear_198_cast_fp16)[name = tensor("op_2944_cast_fp16")]; + tensor var_2946 = const()[name = tensor("op_2946"), val = tensor([1, -1, 20, 64])]; + tensor var_2947_cast_fp16 = reshape(shape = var_2946, x = linear_199_cast_fp16)[name = tensor("op_2947_cast_fp16")]; + tensor value_147_perm_0 = const()[name = tensor("value_147_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_52_y_0_to_fp16 = const()[name = tensor("mul_52_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_52_cast_fp16 = mul(x = var_2941_cast_fp16, y = mul_52_y_0_to_fp16)[name = tensor("mul_52_cast_fp16")]; + tensor matmul_36_transpose_y_0 = const()[name = tensor("matmul_36_transpose_y_0"), val = tensor(true)]; + tensor matmul_36_transpose_x_0 = const()[name = tensor("matmul_36_transpose_x_0"), val = tensor(false)]; + tensor transpose_344_perm_0 = const()[name = tensor("transpose_344_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_345_perm_0 = const()[name = tensor("transpose_345_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_537 = transpose(perm = transpose_345_perm_0, x = var_2944_cast_fp16)[name = tensor("transpose_537")]; + tensor transpose_538 = transpose(perm = transpose_344_perm_0, x = mul_52_cast_fp16)[name = tensor("transpose_538")]; + tensor matmul_36_cast_fp16 = matmul(transpose_x = matmul_36_transpose_x_0, transpose_y = matmul_36_transpose_y_0, x = transpose_538, y = transpose_537)[name = tensor("matmul_36_cast_fp16")]; + tensor softmax_36_axis_0 = const()[name = tensor("softmax_36_axis_0"), val = tensor(-1)]; + tensor softmax_36_cast_fp16 = softmax(axis = softmax_36_axis_0, x = matmul_36_cast_fp16)[name = tensor("softmax_36_cast_fp16")]; + tensor hidden_states_421_transpose_x_0 = const()[name = tensor("hidden_states_421_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_421_transpose_y_0 = const()[name = tensor("hidden_states_421_transpose_y_0"), val = tensor(false)]; + tensor transpose_536 = transpose(perm = value_147_perm_0, x = var_2947_cast_fp16)[name = tensor("transpose_536")]; + tensor hidden_states_421_cast_fp16 = matmul(transpose_x = hidden_states_421_transpose_x_0, transpose_y = hidden_states_421_transpose_y_0, x = softmax_36_cast_fp16, y = transpose_536)[name = tensor("hidden_states_421_cast_fp16")]; + tensor var_2950_perm_0 = const()[name = tensor("op_2950_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_2954 = const()[name = tensor("op_2954"), val = tensor([1, -1, 1280])]; + tensor transpose_535 = transpose(perm = var_2950_perm_0, x = hidden_states_421_cast_fp16)[name = tensor("transpose_535")]; + tensor hidden_states_423_cast_fp16 = reshape(shape = var_2954, x = transpose_535)[name = tensor("hidden_states_423_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(469946816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(471175680))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(471175872)))]; + tensor linear_200_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_423_cast_fp16)[name = tensor("linear_200_cast_fp16")]; + tensor input_517_cast_fp16 = add(x = linear_200_cast_fp16, y = hidden_states_417_cast_fp16)[name = tensor("input_517_cast_fp16")]; + tensor input_519_axes_0 = const()[name = tensor("input_519_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(471178496)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(471181120)))]; + tensor input_519_cast_fp16 = layer_norm(axes = input_519_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_4_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_4_norm2_weight_to_fp16, x = input_517_cast_fp16)[name = tensor("input_519_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(471183744))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(472412608))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_201_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized, x = input_519_cast_fp16)[name = tensor("linear_201_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(472412800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(474378944))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_202_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_202_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(474379136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(476345280))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_203_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_203_cast_fp16")]; + tensor var_2986 = const()[name = tensor("op_2986"), val = tensor([1, -1, 20, 64])]; + tensor var_2987_cast_fp16 = reshape(shape = var_2986, x = linear_201_cast_fp16)[name = tensor("op_2987_cast_fp16")]; + tensor var_2989 = const()[name = tensor("op_2989"), val = tensor([1, -1, 20, 64])]; + tensor var_2990_cast_fp16 = reshape(shape = var_2989, x = linear_202_cast_fp16)[name = tensor("op_2990_cast_fp16")]; + tensor var_2992 = const()[name = tensor("op_2992"), val = tensor([1, -1, 20, 64])]; + tensor var_2993_cast_fp16 = reshape(shape = var_2992, x = linear_203_cast_fp16)[name = tensor("op_2993_cast_fp16")]; + tensor value_151_perm_0 = const()[name = tensor("value_151_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_53_y_0_to_fp16 = const()[name = tensor("mul_53_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_53_cast_fp16 = mul(x = var_2987_cast_fp16, y = mul_53_y_0_to_fp16)[name = tensor("mul_53_cast_fp16")]; + tensor matmul_37_transpose_y_0 = const()[name = tensor("matmul_37_transpose_y_0"), val = tensor(true)]; + tensor matmul_37_transpose_x_0 = const()[name = tensor("matmul_37_transpose_x_0"), val = tensor(false)]; + tensor transpose_346_perm_0 = const()[name = tensor("transpose_346_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_347_perm_0 = const()[name = tensor("transpose_347_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_533 = transpose(perm = transpose_347_perm_0, x = var_2990_cast_fp16)[name = tensor("transpose_533")]; + tensor transpose_534 = transpose(perm = transpose_346_perm_0, x = mul_53_cast_fp16)[name = tensor("transpose_534")]; + tensor matmul_37_cast_fp16 = matmul(transpose_x = matmul_37_transpose_x_0, transpose_y = matmul_37_transpose_y_0, x = transpose_534, y = transpose_533)[name = tensor("matmul_37_cast_fp16")]; + tensor softmax_37_axis_0 = const()[name = tensor("softmax_37_axis_0"), val = tensor(-1)]; + tensor softmax_37_cast_fp16 = softmax(axis = softmax_37_axis_0, x = matmul_37_cast_fp16)[name = tensor("softmax_37_cast_fp16")]; + tensor hidden_states_427_transpose_x_0 = const()[name = tensor("hidden_states_427_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_427_transpose_y_0 = const()[name = tensor("hidden_states_427_transpose_y_0"), val = tensor(false)]; + tensor transpose_532 = transpose(perm = value_151_perm_0, x = var_2993_cast_fp16)[name = tensor("transpose_532")]; + tensor hidden_states_427_cast_fp16 = matmul(transpose_x = hidden_states_427_transpose_x_0, transpose_y = hidden_states_427_transpose_y_0, x = softmax_37_cast_fp16, y = transpose_532)[name = tensor("hidden_states_427_cast_fp16")]; + tensor var_2996_perm_0 = const()[name = tensor("op_2996_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3000 = const()[name = tensor("op_3000"), val = tensor([1, -1, 1280])]; + tensor transpose_531 = transpose(perm = var_2996_perm_0, x = hidden_states_427_cast_fp16)[name = tensor("transpose_531")]; + tensor hidden_states_429_cast_fp16 = reshape(shape = var_3000, x = transpose_531)[name = tensor("hidden_states_429_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(476345472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(477574336))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(477574528)))]; + tensor linear_204_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_429_cast_fp16)[name = tensor("linear_204_cast_fp16")]; + tensor input_525_cast_fp16 = add(x = linear_204_cast_fp16, y = input_517_cast_fp16)[name = tensor("input_525_cast_fp16")]; + tensor input_527_axes_0 = const()[name = tensor("input_527_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(477577152)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(477579776)))]; + tensor input_527_cast_fp16 = layer_norm(axes = input_527_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_4_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_4_norm3_weight_to_fp16, x = input_525_cast_fp16)[name = tensor("input_527_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(477582400))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(487412864))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(487413056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(487420800))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_205_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized, x = input_527_cast_fp16)[name = tensor("linear_205_cast_fp16")]; + tensor var_3022_split_sizes_0 = const()[name = tensor("op_3022_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3022_axis_0 = const()[name = tensor("op_3022_axis_0"), val = tensor(-1)]; + tensor var_3022_cast_fp16_0, tensor var_3022_cast_fp16_1 = split(axis = var_3022_axis_0, split_sizes = var_3022_split_sizes_0, x = linear_205_cast_fp16)[name = tensor("op_3022_cast_fp16")]; + tensor var_3024_mode_0 = const()[name = tensor("op_3024_mode_0"), val = tensor("EXACT")]; + tensor var_3024_cast_fp16 = gelu(mode = var_3024_mode_0, x = var_3022_cast_fp16_1)[name = tensor("op_3024_cast_fp16")]; + tensor input_529_cast_fp16 = mul(x = var_3022_cast_fp16_0, y = var_3024_cast_fp16)[name = tensor("input_529_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(487420992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(492336256))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(492336448)))]; + tensor linear_206_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized, x = input_529_cast_fp16)[name = tensor("linear_206_cast_fp16")]; + tensor hidden_states_437_cast_fp16 = add(x = linear_206_cast_fp16, y = input_525_cast_fp16)[name = tensor("hidden_states_437_cast_fp16")]; + tensor hidden_states_439_axes_0 = const()[name = tensor("hidden_states_439_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(492339072)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(492341696)))]; + tensor hidden_states_439_cast_fp16 = layer_norm(axes = hidden_states_439_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_5_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_5_norm1_weight_to_fp16, x = hidden_states_437_cast_fp16)[name = tensor("hidden_states_439_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(492344320))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(493573184))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_207_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_439_cast_fp16)[name = tensor("linear_207_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(493573376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(494802240))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_208_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_439_cast_fp16)[name = tensor("linear_208_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(494802432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(496031296))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_209_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_439_cast_fp16)[name = tensor("linear_209_cast_fp16")]; + tensor var_3059 = const()[name = tensor("op_3059"), val = tensor([1, -1, 20, 64])]; + tensor var_3060_cast_fp16 = reshape(shape = var_3059, x = linear_207_cast_fp16)[name = tensor("op_3060_cast_fp16")]; + tensor var_3062 = const()[name = tensor("op_3062"), val = tensor([1, -1, 20, 64])]; + tensor var_3063_cast_fp16 = reshape(shape = var_3062, x = linear_208_cast_fp16)[name = tensor("op_3063_cast_fp16")]; + tensor var_3065 = const()[name = tensor("op_3065"), val = tensor([1, -1, 20, 64])]; + tensor var_3066_cast_fp16 = reshape(shape = var_3065, x = linear_209_cast_fp16)[name = tensor("op_3066_cast_fp16")]; + tensor value_155_perm_0 = const()[name = tensor("value_155_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_54_y_0_to_fp16 = const()[name = tensor("mul_54_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_54_cast_fp16 = mul(x = var_3060_cast_fp16, y = mul_54_y_0_to_fp16)[name = tensor("mul_54_cast_fp16")]; + tensor matmul_38_transpose_y_0 = const()[name = tensor("matmul_38_transpose_y_0"), val = tensor(true)]; + tensor matmul_38_transpose_x_0 = const()[name = tensor("matmul_38_transpose_x_0"), val = tensor(false)]; + tensor transpose_348_perm_0 = const()[name = tensor("transpose_348_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_349_perm_0 = const()[name = tensor("transpose_349_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_529 = transpose(perm = transpose_349_perm_0, x = var_3063_cast_fp16)[name = tensor("transpose_529")]; + tensor transpose_530 = transpose(perm = transpose_348_perm_0, x = mul_54_cast_fp16)[name = tensor("transpose_530")]; + tensor matmul_38_cast_fp16 = matmul(transpose_x = matmul_38_transpose_x_0, transpose_y = matmul_38_transpose_y_0, x = transpose_530, y = transpose_529)[name = tensor("matmul_38_cast_fp16")]; + tensor softmax_38_axis_0 = const()[name = tensor("softmax_38_axis_0"), val = tensor(-1)]; + tensor softmax_38_cast_fp16 = softmax(axis = softmax_38_axis_0, x = matmul_38_cast_fp16)[name = tensor("softmax_38_cast_fp16")]; + tensor hidden_states_441_transpose_x_0 = const()[name = tensor("hidden_states_441_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_441_transpose_y_0 = const()[name = tensor("hidden_states_441_transpose_y_0"), val = tensor(false)]; + tensor transpose_528 = transpose(perm = value_155_perm_0, x = var_3066_cast_fp16)[name = tensor("transpose_528")]; + tensor hidden_states_441_cast_fp16 = matmul(transpose_x = hidden_states_441_transpose_x_0, transpose_y = hidden_states_441_transpose_y_0, x = softmax_38_cast_fp16, y = transpose_528)[name = tensor("hidden_states_441_cast_fp16")]; + tensor var_3069_perm_0 = const()[name = tensor("op_3069_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3073 = const()[name = tensor("op_3073"), val = tensor([1, -1, 1280])]; + tensor transpose_527 = transpose(perm = var_3069_perm_0, x = hidden_states_441_cast_fp16)[name = tensor("transpose_527")]; + tensor hidden_states_443_cast_fp16 = reshape(shape = var_3073, x = transpose_527)[name = tensor("hidden_states_443_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(496031488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(497260352))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(497260544)))]; + tensor linear_210_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_443_cast_fp16)[name = tensor("linear_210_cast_fp16")]; + tensor input_537_cast_fp16 = add(x = linear_210_cast_fp16, y = hidden_states_437_cast_fp16)[name = tensor("input_537_cast_fp16")]; + tensor input_539_axes_0 = const()[name = tensor("input_539_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(497263168)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(497265792)))]; + tensor input_539_cast_fp16 = layer_norm(axes = input_539_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_5_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_5_norm2_weight_to_fp16, x = input_537_cast_fp16)[name = tensor("input_539_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(497268416))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(498497280))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_211_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized, x = input_539_cast_fp16)[name = tensor("linear_211_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(498497472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(500463616))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_212_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_212_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(500463808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(502429952))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_213_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_213_cast_fp16")]; + tensor var_3105 = const()[name = tensor("op_3105"), val = tensor([1, -1, 20, 64])]; + tensor var_3106_cast_fp16 = reshape(shape = var_3105, x = linear_211_cast_fp16)[name = tensor("op_3106_cast_fp16")]; + tensor var_3108 = const()[name = tensor("op_3108"), val = tensor([1, -1, 20, 64])]; + tensor var_3109_cast_fp16 = reshape(shape = var_3108, x = linear_212_cast_fp16)[name = tensor("op_3109_cast_fp16")]; + tensor var_3111 = const()[name = tensor("op_3111"), val = tensor([1, -1, 20, 64])]; + tensor var_3112_cast_fp16 = reshape(shape = var_3111, x = linear_213_cast_fp16)[name = tensor("op_3112_cast_fp16")]; + tensor value_159_perm_0 = const()[name = tensor("value_159_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_55_y_0_to_fp16 = const()[name = tensor("mul_55_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_55_cast_fp16 = mul(x = var_3106_cast_fp16, y = mul_55_y_0_to_fp16)[name = tensor("mul_55_cast_fp16")]; + tensor matmul_39_transpose_y_0 = const()[name = tensor("matmul_39_transpose_y_0"), val = tensor(true)]; + tensor matmul_39_transpose_x_0 = const()[name = tensor("matmul_39_transpose_x_0"), val = tensor(false)]; + tensor transpose_350_perm_0 = const()[name = tensor("transpose_350_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_351_perm_0 = const()[name = tensor("transpose_351_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_525 = transpose(perm = transpose_351_perm_0, x = var_3109_cast_fp16)[name = tensor("transpose_525")]; + tensor transpose_526 = transpose(perm = transpose_350_perm_0, x = mul_55_cast_fp16)[name = tensor("transpose_526")]; + tensor matmul_39_cast_fp16 = matmul(transpose_x = matmul_39_transpose_x_0, transpose_y = matmul_39_transpose_y_0, x = transpose_526, y = transpose_525)[name = tensor("matmul_39_cast_fp16")]; + tensor softmax_39_axis_0 = const()[name = tensor("softmax_39_axis_0"), val = tensor(-1)]; + tensor softmax_39_cast_fp16 = softmax(axis = softmax_39_axis_0, x = matmul_39_cast_fp16)[name = tensor("softmax_39_cast_fp16")]; + tensor hidden_states_447_transpose_x_0 = const()[name = tensor("hidden_states_447_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_447_transpose_y_0 = const()[name = tensor("hidden_states_447_transpose_y_0"), val = tensor(false)]; + tensor transpose_524 = transpose(perm = value_159_perm_0, x = var_3112_cast_fp16)[name = tensor("transpose_524")]; + tensor hidden_states_447_cast_fp16 = matmul(transpose_x = hidden_states_447_transpose_x_0, transpose_y = hidden_states_447_transpose_y_0, x = softmax_39_cast_fp16, y = transpose_524)[name = tensor("hidden_states_447_cast_fp16")]; + tensor var_3115_perm_0 = const()[name = tensor("op_3115_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3119 = const()[name = tensor("op_3119"), val = tensor([1, -1, 1280])]; + tensor transpose_523 = transpose(perm = var_3115_perm_0, x = hidden_states_447_cast_fp16)[name = tensor("transpose_523")]; + tensor hidden_states_449_cast_fp16 = reshape(shape = var_3119, x = transpose_523)[name = tensor("hidden_states_449_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(502430144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(503659008))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(503659200)))]; + tensor linear_214_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_449_cast_fp16)[name = tensor("linear_214_cast_fp16")]; + tensor input_545_cast_fp16 = add(x = linear_214_cast_fp16, y = input_537_cast_fp16)[name = tensor("input_545_cast_fp16")]; + tensor input_547_axes_0 = const()[name = tensor("input_547_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(503661824)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(503664448)))]; + tensor input_547_cast_fp16 = layer_norm(axes = input_547_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_5_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_5_norm3_weight_to_fp16, x = input_545_cast_fp16)[name = tensor("input_547_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(503667072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(513497536))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(513497728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(513505472))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_215_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized, x = input_547_cast_fp16)[name = tensor("linear_215_cast_fp16")]; + tensor var_3141_split_sizes_0 = const()[name = tensor("op_3141_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3141_axis_0 = const()[name = tensor("op_3141_axis_0"), val = tensor(-1)]; + tensor var_3141_cast_fp16_0, tensor var_3141_cast_fp16_1 = split(axis = var_3141_axis_0, split_sizes = var_3141_split_sizes_0, x = linear_215_cast_fp16)[name = tensor("op_3141_cast_fp16")]; + tensor var_3143_mode_0 = const()[name = tensor("op_3143_mode_0"), val = tensor("EXACT")]; + tensor var_3143_cast_fp16 = gelu(mode = var_3143_mode_0, x = var_3141_cast_fp16_1)[name = tensor("op_3143_cast_fp16")]; + tensor input_549_cast_fp16 = mul(x = var_3141_cast_fp16_0, y = var_3143_cast_fp16)[name = tensor("input_549_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(513505664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(518420928))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(518421120)))]; + tensor linear_216_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized, x = input_549_cast_fp16)[name = tensor("linear_216_cast_fp16")]; + tensor hidden_states_457_cast_fp16 = add(x = linear_216_cast_fp16, y = input_545_cast_fp16)[name = tensor("hidden_states_457_cast_fp16")]; + tensor hidden_states_459_axes_0 = const()[name = tensor("hidden_states_459_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(518423744)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(518426368)))]; + tensor hidden_states_459_cast_fp16 = layer_norm(axes = hidden_states_459_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_6_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_6_norm1_weight_to_fp16, x = hidden_states_457_cast_fp16)[name = tensor("hidden_states_459_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(518428992))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(519657856))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_217_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_459_cast_fp16)[name = tensor("linear_217_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(519658048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(520886912))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_218_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_459_cast_fp16)[name = tensor("linear_218_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(520887104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(522115968))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_219_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_459_cast_fp16)[name = tensor("linear_219_cast_fp16")]; + tensor var_3178 = const()[name = tensor("op_3178"), val = tensor([1, -1, 20, 64])]; + tensor var_3179_cast_fp16 = reshape(shape = var_3178, x = linear_217_cast_fp16)[name = tensor("op_3179_cast_fp16")]; + tensor var_3181 = const()[name = tensor("op_3181"), val = tensor([1, -1, 20, 64])]; + tensor var_3182_cast_fp16 = reshape(shape = var_3181, x = linear_218_cast_fp16)[name = tensor("op_3182_cast_fp16")]; + tensor var_3184 = const()[name = tensor("op_3184"), val = tensor([1, -1, 20, 64])]; + tensor var_3185_cast_fp16 = reshape(shape = var_3184, x = linear_219_cast_fp16)[name = tensor("op_3185_cast_fp16")]; + tensor value_163_perm_0 = const()[name = tensor("value_163_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_56_y_0_to_fp16 = const()[name = tensor("mul_56_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_56_cast_fp16 = mul(x = var_3179_cast_fp16, y = mul_56_y_0_to_fp16)[name = tensor("mul_56_cast_fp16")]; + tensor matmul_40_transpose_y_0 = const()[name = tensor("matmul_40_transpose_y_0"), val = tensor(true)]; + tensor matmul_40_transpose_x_0 = const()[name = tensor("matmul_40_transpose_x_0"), val = tensor(false)]; + tensor transpose_352_perm_0 = const()[name = tensor("transpose_352_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_353_perm_0 = const()[name = tensor("transpose_353_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_521 = transpose(perm = transpose_353_perm_0, x = var_3182_cast_fp16)[name = tensor("transpose_521")]; + tensor transpose_522 = transpose(perm = transpose_352_perm_0, x = mul_56_cast_fp16)[name = tensor("transpose_522")]; + tensor matmul_40_cast_fp16 = matmul(transpose_x = matmul_40_transpose_x_0, transpose_y = matmul_40_transpose_y_0, x = transpose_522, y = transpose_521)[name = tensor("matmul_40_cast_fp16")]; + tensor softmax_40_axis_0 = const()[name = tensor("softmax_40_axis_0"), val = tensor(-1)]; + tensor softmax_40_cast_fp16 = softmax(axis = softmax_40_axis_0, x = matmul_40_cast_fp16)[name = tensor("softmax_40_cast_fp16")]; + tensor hidden_states_461_transpose_x_0 = const()[name = tensor("hidden_states_461_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_461_transpose_y_0 = const()[name = tensor("hidden_states_461_transpose_y_0"), val = tensor(false)]; + tensor transpose_520 = transpose(perm = value_163_perm_0, x = var_3185_cast_fp16)[name = tensor("transpose_520")]; + tensor hidden_states_461_cast_fp16 = matmul(transpose_x = hidden_states_461_transpose_x_0, transpose_y = hidden_states_461_transpose_y_0, x = softmax_40_cast_fp16, y = transpose_520)[name = tensor("hidden_states_461_cast_fp16")]; + tensor var_3188_perm_0 = const()[name = tensor("op_3188_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3192 = const()[name = tensor("op_3192"), val = tensor([1, -1, 1280])]; + tensor transpose_519 = transpose(perm = var_3188_perm_0, x = hidden_states_461_cast_fp16)[name = tensor("transpose_519")]; + tensor hidden_states_463_cast_fp16 = reshape(shape = var_3192, x = transpose_519)[name = tensor("hidden_states_463_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(522116160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(523345024))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(523345216)))]; + tensor linear_220_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_463_cast_fp16)[name = tensor("linear_220_cast_fp16")]; + tensor input_557_cast_fp16 = add(x = linear_220_cast_fp16, y = hidden_states_457_cast_fp16)[name = tensor("input_557_cast_fp16")]; + tensor input_559_axes_0 = const()[name = tensor("input_559_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(523347840)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(523350464)))]; + tensor input_559_cast_fp16 = layer_norm(axes = input_559_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_6_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_6_norm2_weight_to_fp16, x = input_557_cast_fp16)[name = tensor("input_559_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(523353088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(524581952))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_221_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized, x = input_559_cast_fp16)[name = tensor("linear_221_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(524582144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(526548288))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_222_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_222_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(526548480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(528514624))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_223_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_223_cast_fp16")]; + tensor var_3224 = const()[name = tensor("op_3224"), val = tensor([1, -1, 20, 64])]; + tensor var_3225_cast_fp16 = reshape(shape = var_3224, x = linear_221_cast_fp16)[name = tensor("op_3225_cast_fp16")]; + tensor var_3227 = const()[name = tensor("op_3227"), val = tensor([1, -1, 20, 64])]; + tensor var_3228_cast_fp16 = reshape(shape = var_3227, x = linear_222_cast_fp16)[name = tensor("op_3228_cast_fp16")]; + tensor var_3230 = const()[name = tensor("op_3230"), val = tensor([1, -1, 20, 64])]; + tensor var_3231_cast_fp16 = reshape(shape = var_3230, x = linear_223_cast_fp16)[name = tensor("op_3231_cast_fp16")]; + tensor value_167_perm_0 = const()[name = tensor("value_167_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_57_y_0_to_fp16 = const()[name = tensor("mul_57_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_57_cast_fp16 = mul(x = var_3225_cast_fp16, y = mul_57_y_0_to_fp16)[name = tensor("mul_57_cast_fp16")]; + tensor matmul_41_transpose_y_0 = const()[name = tensor("matmul_41_transpose_y_0"), val = tensor(true)]; + tensor matmul_41_transpose_x_0 = const()[name = tensor("matmul_41_transpose_x_0"), val = tensor(false)]; + tensor transpose_354_perm_0 = const()[name = tensor("transpose_354_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_355_perm_0 = const()[name = tensor("transpose_355_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_517 = transpose(perm = transpose_355_perm_0, x = var_3228_cast_fp16)[name = tensor("transpose_517")]; + tensor transpose_518 = transpose(perm = transpose_354_perm_0, x = mul_57_cast_fp16)[name = tensor("transpose_518")]; + tensor matmul_41_cast_fp16 = matmul(transpose_x = matmul_41_transpose_x_0, transpose_y = matmul_41_transpose_y_0, x = transpose_518, y = transpose_517)[name = tensor("matmul_41_cast_fp16")]; + tensor softmax_41_axis_0 = const()[name = tensor("softmax_41_axis_0"), val = tensor(-1)]; + tensor softmax_41_cast_fp16 = softmax(axis = softmax_41_axis_0, x = matmul_41_cast_fp16)[name = tensor("softmax_41_cast_fp16")]; + tensor hidden_states_467_transpose_x_0 = const()[name = tensor("hidden_states_467_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_467_transpose_y_0 = const()[name = tensor("hidden_states_467_transpose_y_0"), val = tensor(false)]; + tensor transpose_516 = transpose(perm = value_167_perm_0, x = var_3231_cast_fp16)[name = tensor("transpose_516")]; + tensor hidden_states_467_cast_fp16 = matmul(transpose_x = hidden_states_467_transpose_x_0, transpose_y = hidden_states_467_transpose_y_0, x = softmax_41_cast_fp16, y = transpose_516)[name = tensor("hidden_states_467_cast_fp16")]; + tensor var_3234_perm_0 = const()[name = tensor("op_3234_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3238 = const()[name = tensor("op_3238"), val = tensor([1, -1, 1280])]; + tensor transpose_515 = transpose(perm = var_3234_perm_0, x = hidden_states_467_cast_fp16)[name = tensor("transpose_515")]; + tensor hidden_states_469_cast_fp16 = reshape(shape = var_3238, x = transpose_515)[name = tensor("hidden_states_469_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(528514816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529743680))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529743872)))]; + tensor linear_224_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_469_cast_fp16)[name = tensor("linear_224_cast_fp16")]; + tensor input_565_cast_fp16 = add(x = linear_224_cast_fp16, y = input_557_cast_fp16)[name = tensor("input_565_cast_fp16")]; + tensor input_567_axes_0 = const()[name = tensor("input_567_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529746496)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529749120)))]; + tensor input_567_cast_fp16 = layer_norm(axes = input_567_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_6_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_6_norm3_weight_to_fp16, x = input_565_cast_fp16)[name = tensor("input_567_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(529751744))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(539582208))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(539582400))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(539590144))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_225_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized, x = input_567_cast_fp16)[name = tensor("linear_225_cast_fp16")]; + tensor var_3260_split_sizes_0 = const()[name = tensor("op_3260_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3260_axis_0 = const()[name = tensor("op_3260_axis_0"), val = tensor(-1)]; + tensor var_3260_cast_fp16_0, tensor var_3260_cast_fp16_1 = split(axis = var_3260_axis_0, split_sizes = var_3260_split_sizes_0, x = linear_225_cast_fp16)[name = tensor("op_3260_cast_fp16")]; + tensor var_3262_mode_0 = const()[name = tensor("op_3262_mode_0"), val = tensor("EXACT")]; + tensor var_3262_cast_fp16 = gelu(mode = var_3262_mode_0, x = var_3260_cast_fp16_1)[name = tensor("op_3262_cast_fp16")]; + tensor input_569_cast_fp16 = mul(x = var_3260_cast_fp16_0, y = var_3262_cast_fp16)[name = tensor("input_569_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(539590336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(544505600))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(544505792)))]; + tensor linear_226_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized, x = input_569_cast_fp16)[name = tensor("linear_226_cast_fp16")]; + tensor hidden_states_477_cast_fp16 = add(x = linear_226_cast_fp16, y = input_565_cast_fp16)[name = tensor("hidden_states_477_cast_fp16")]; + tensor hidden_states_479_axes_0 = const()[name = tensor("hidden_states_479_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(544508416)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(544511040)))]; + tensor hidden_states_479_cast_fp16 = layer_norm(axes = hidden_states_479_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_7_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_7_norm1_weight_to_fp16, x = hidden_states_477_cast_fp16)[name = tensor("hidden_states_479_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(544513664))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(545742528))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_227_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_479_cast_fp16)[name = tensor("linear_227_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(545742720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(546971584))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_228_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_479_cast_fp16)[name = tensor("linear_228_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(546971776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(548200640))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_229_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_479_cast_fp16)[name = tensor("linear_229_cast_fp16")]; + tensor var_3297 = const()[name = tensor("op_3297"), val = tensor([1, -1, 20, 64])]; + tensor var_3298_cast_fp16 = reshape(shape = var_3297, x = linear_227_cast_fp16)[name = tensor("op_3298_cast_fp16")]; + tensor var_3300 = const()[name = tensor("op_3300"), val = tensor([1, -1, 20, 64])]; + tensor var_3301_cast_fp16 = reshape(shape = var_3300, x = linear_228_cast_fp16)[name = tensor("op_3301_cast_fp16")]; + tensor var_3303 = const()[name = tensor("op_3303"), val = tensor([1, -1, 20, 64])]; + tensor var_3304_cast_fp16 = reshape(shape = var_3303, x = linear_229_cast_fp16)[name = tensor("op_3304_cast_fp16")]; + tensor value_171_perm_0 = const()[name = tensor("value_171_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_58_y_0_to_fp16 = const()[name = tensor("mul_58_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_58_cast_fp16 = mul(x = var_3298_cast_fp16, y = mul_58_y_0_to_fp16)[name = tensor("mul_58_cast_fp16")]; + tensor matmul_42_transpose_y_0 = const()[name = tensor("matmul_42_transpose_y_0"), val = tensor(true)]; + tensor matmul_42_transpose_x_0 = const()[name = tensor("matmul_42_transpose_x_0"), val = tensor(false)]; + tensor transpose_356_perm_0 = const()[name = tensor("transpose_356_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_357_perm_0 = const()[name = tensor("transpose_357_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_513 = transpose(perm = transpose_357_perm_0, x = var_3301_cast_fp16)[name = tensor("transpose_513")]; + tensor transpose_514 = transpose(perm = transpose_356_perm_0, x = mul_58_cast_fp16)[name = tensor("transpose_514")]; + tensor matmul_42_cast_fp16 = matmul(transpose_x = matmul_42_transpose_x_0, transpose_y = matmul_42_transpose_y_0, x = transpose_514, y = transpose_513)[name = tensor("matmul_42_cast_fp16")]; + tensor softmax_42_axis_0 = const()[name = tensor("softmax_42_axis_0"), val = tensor(-1)]; + tensor softmax_42_cast_fp16 = softmax(axis = softmax_42_axis_0, x = matmul_42_cast_fp16)[name = tensor("softmax_42_cast_fp16")]; + tensor hidden_states_481_transpose_x_0 = const()[name = tensor("hidden_states_481_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_481_transpose_y_0 = const()[name = tensor("hidden_states_481_transpose_y_0"), val = tensor(false)]; + tensor transpose_512 = transpose(perm = value_171_perm_0, x = var_3304_cast_fp16)[name = tensor("transpose_512")]; + tensor hidden_states_481_cast_fp16 = matmul(transpose_x = hidden_states_481_transpose_x_0, transpose_y = hidden_states_481_transpose_y_0, x = softmax_42_cast_fp16, y = transpose_512)[name = tensor("hidden_states_481_cast_fp16")]; + tensor var_3307_perm_0 = const()[name = tensor("op_3307_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3311 = const()[name = tensor("op_3311"), val = tensor([1, -1, 1280])]; + tensor transpose_511 = transpose(perm = var_3307_perm_0, x = hidden_states_481_cast_fp16)[name = tensor("transpose_511")]; + tensor hidden_states_483_cast_fp16 = reshape(shape = var_3311, x = transpose_511)[name = tensor("hidden_states_483_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(548200832))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(549429696))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(549429888)))]; + tensor linear_230_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_483_cast_fp16)[name = tensor("linear_230_cast_fp16")]; + tensor input_577_cast_fp16 = add(x = linear_230_cast_fp16, y = hidden_states_477_cast_fp16)[name = tensor("input_577_cast_fp16")]; + tensor input_579_axes_0 = const()[name = tensor("input_579_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(549432512)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(549435136)))]; + tensor input_579_cast_fp16 = layer_norm(axes = input_579_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_7_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_7_norm2_weight_to_fp16, x = input_577_cast_fp16)[name = tensor("input_579_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(549437760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(550666624))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_231_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized, x = input_579_cast_fp16)[name = tensor("linear_231_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(550666816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(552632960))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_232_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_232_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(552633152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(554599296))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_233_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_233_cast_fp16")]; + tensor var_3343 = const()[name = tensor("op_3343"), val = tensor([1, -1, 20, 64])]; + tensor var_3344_cast_fp16 = reshape(shape = var_3343, x = linear_231_cast_fp16)[name = tensor("op_3344_cast_fp16")]; + tensor var_3346 = const()[name = tensor("op_3346"), val = tensor([1, -1, 20, 64])]; + tensor var_3347_cast_fp16 = reshape(shape = var_3346, x = linear_232_cast_fp16)[name = tensor("op_3347_cast_fp16")]; + tensor var_3349 = const()[name = tensor("op_3349"), val = tensor([1, -1, 20, 64])]; + tensor var_3350_cast_fp16 = reshape(shape = var_3349, x = linear_233_cast_fp16)[name = tensor("op_3350_cast_fp16")]; + tensor value_175_perm_0 = const()[name = tensor("value_175_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_59_y_0_to_fp16 = const()[name = tensor("mul_59_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_59_cast_fp16 = mul(x = var_3344_cast_fp16, y = mul_59_y_0_to_fp16)[name = tensor("mul_59_cast_fp16")]; + tensor matmul_43_transpose_y_0 = const()[name = tensor("matmul_43_transpose_y_0"), val = tensor(true)]; + tensor matmul_43_transpose_x_0 = const()[name = tensor("matmul_43_transpose_x_0"), val = tensor(false)]; + tensor transpose_358_perm_0 = const()[name = tensor("transpose_358_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_359_perm_0 = const()[name = tensor("transpose_359_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_509 = transpose(perm = transpose_359_perm_0, x = var_3347_cast_fp16)[name = tensor("transpose_509")]; + tensor transpose_510 = transpose(perm = transpose_358_perm_0, x = mul_59_cast_fp16)[name = tensor("transpose_510")]; + tensor matmul_43_cast_fp16 = matmul(transpose_x = matmul_43_transpose_x_0, transpose_y = matmul_43_transpose_y_0, x = transpose_510, y = transpose_509)[name = tensor("matmul_43_cast_fp16")]; + tensor softmax_43_axis_0 = const()[name = tensor("softmax_43_axis_0"), val = tensor(-1)]; + tensor softmax_43_cast_fp16 = softmax(axis = softmax_43_axis_0, x = matmul_43_cast_fp16)[name = tensor("softmax_43_cast_fp16")]; + tensor hidden_states_487_transpose_x_0 = const()[name = tensor("hidden_states_487_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_487_transpose_y_0 = const()[name = tensor("hidden_states_487_transpose_y_0"), val = tensor(false)]; + tensor transpose_508 = transpose(perm = value_175_perm_0, x = var_3350_cast_fp16)[name = tensor("transpose_508")]; + tensor hidden_states_487_cast_fp16 = matmul(transpose_x = hidden_states_487_transpose_x_0, transpose_y = hidden_states_487_transpose_y_0, x = softmax_43_cast_fp16, y = transpose_508)[name = tensor("hidden_states_487_cast_fp16")]; + tensor var_3353_perm_0 = const()[name = tensor("op_3353_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3357 = const()[name = tensor("op_3357"), val = tensor([1, -1, 1280])]; + tensor transpose_507 = transpose(perm = var_3353_perm_0, x = hidden_states_487_cast_fp16)[name = tensor("transpose_507")]; + tensor hidden_states_489_cast_fp16 = reshape(shape = var_3357, x = transpose_507)[name = tensor("hidden_states_489_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(554599488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(555828352))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(555828544)))]; + tensor linear_234_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_489_cast_fp16)[name = tensor("linear_234_cast_fp16")]; + tensor input_585_cast_fp16 = add(x = linear_234_cast_fp16, y = input_577_cast_fp16)[name = tensor("input_585_cast_fp16")]; + tensor input_587_axes_0 = const()[name = tensor("input_587_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(555831168)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(555833792)))]; + tensor input_587_cast_fp16 = layer_norm(axes = input_587_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_7_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_7_norm3_weight_to_fp16, x = input_585_cast_fp16)[name = tensor("input_587_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(555836416))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(565666880))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(565667072))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(565674816))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_235_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized, x = input_587_cast_fp16)[name = tensor("linear_235_cast_fp16")]; + tensor var_3379_split_sizes_0 = const()[name = tensor("op_3379_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3379_axis_0 = const()[name = tensor("op_3379_axis_0"), val = tensor(-1)]; + tensor var_3379_cast_fp16_0, tensor var_3379_cast_fp16_1 = split(axis = var_3379_axis_0, split_sizes = var_3379_split_sizes_0, x = linear_235_cast_fp16)[name = tensor("op_3379_cast_fp16")]; + tensor var_3381_mode_0 = const()[name = tensor("op_3381_mode_0"), val = tensor("EXACT")]; + tensor var_3381_cast_fp16 = gelu(mode = var_3381_mode_0, x = var_3379_cast_fp16_1)[name = tensor("op_3381_cast_fp16")]; + tensor input_589_cast_fp16 = mul(x = var_3379_cast_fp16_0, y = var_3381_cast_fp16)[name = tensor("input_589_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(565675008))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(570590272))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(570590464)))]; + tensor linear_236_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized, x = input_589_cast_fp16)[name = tensor("linear_236_cast_fp16")]; + tensor hidden_states_497_cast_fp16 = add(x = linear_236_cast_fp16, y = input_585_cast_fp16)[name = tensor("hidden_states_497_cast_fp16")]; + tensor hidden_states_499_axes_0 = const()[name = tensor("hidden_states_499_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(570593088)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(570595712)))]; + tensor hidden_states_499_cast_fp16 = layer_norm(axes = hidden_states_499_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_8_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_8_norm1_weight_to_fp16, x = hidden_states_497_cast_fp16)[name = tensor("hidden_states_499_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(570598336))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(571827200))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_237_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_499_cast_fp16)[name = tensor("linear_237_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(571827392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(573056256))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_238_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_499_cast_fp16)[name = tensor("linear_238_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(573056448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(574285312))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_239_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_499_cast_fp16)[name = tensor("linear_239_cast_fp16")]; + tensor var_3416 = const()[name = tensor("op_3416"), val = tensor([1, -1, 20, 64])]; + tensor var_3417_cast_fp16 = reshape(shape = var_3416, x = linear_237_cast_fp16)[name = tensor("op_3417_cast_fp16")]; + tensor var_3419 = const()[name = tensor("op_3419"), val = tensor([1, -1, 20, 64])]; + tensor var_3420_cast_fp16 = reshape(shape = var_3419, x = linear_238_cast_fp16)[name = tensor("op_3420_cast_fp16")]; + tensor var_3422 = const()[name = tensor("op_3422"), val = tensor([1, -1, 20, 64])]; + tensor var_3423_cast_fp16 = reshape(shape = var_3422, x = linear_239_cast_fp16)[name = tensor("op_3423_cast_fp16")]; + tensor value_179_perm_0 = const()[name = tensor("value_179_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_60_y_0_to_fp16 = const()[name = tensor("mul_60_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_60_cast_fp16 = mul(x = var_3417_cast_fp16, y = mul_60_y_0_to_fp16)[name = tensor("mul_60_cast_fp16")]; + tensor matmul_44_transpose_y_0 = const()[name = tensor("matmul_44_transpose_y_0"), val = tensor(true)]; + tensor matmul_44_transpose_x_0 = const()[name = tensor("matmul_44_transpose_x_0"), val = tensor(false)]; + tensor transpose_360_perm_0 = const()[name = tensor("transpose_360_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_361_perm_0 = const()[name = tensor("transpose_361_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_505 = transpose(perm = transpose_361_perm_0, x = var_3420_cast_fp16)[name = tensor("transpose_505")]; + tensor transpose_506 = transpose(perm = transpose_360_perm_0, x = mul_60_cast_fp16)[name = tensor("transpose_506")]; + tensor matmul_44_cast_fp16 = matmul(transpose_x = matmul_44_transpose_x_0, transpose_y = matmul_44_transpose_y_0, x = transpose_506, y = transpose_505)[name = tensor("matmul_44_cast_fp16")]; + tensor softmax_44_axis_0 = const()[name = tensor("softmax_44_axis_0"), val = tensor(-1)]; + tensor softmax_44_cast_fp16 = softmax(axis = softmax_44_axis_0, x = matmul_44_cast_fp16)[name = tensor("softmax_44_cast_fp16")]; + tensor hidden_states_501_transpose_x_0 = const()[name = tensor("hidden_states_501_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_501_transpose_y_0 = const()[name = tensor("hidden_states_501_transpose_y_0"), val = tensor(false)]; + tensor transpose_504 = transpose(perm = value_179_perm_0, x = var_3423_cast_fp16)[name = tensor("transpose_504")]; + tensor hidden_states_501_cast_fp16 = matmul(transpose_x = hidden_states_501_transpose_x_0, transpose_y = hidden_states_501_transpose_y_0, x = softmax_44_cast_fp16, y = transpose_504)[name = tensor("hidden_states_501_cast_fp16")]; + tensor var_3426_perm_0 = const()[name = tensor("op_3426_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3430 = const()[name = tensor("op_3430"), val = tensor([1, -1, 1280])]; + tensor transpose_503 = transpose(perm = var_3426_perm_0, x = hidden_states_501_cast_fp16)[name = tensor("transpose_503")]; + tensor hidden_states_503_cast_fp16 = reshape(shape = var_3430, x = transpose_503)[name = tensor("hidden_states_503_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(574285504))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(575514368))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(575514560)))]; + tensor linear_240_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_503_cast_fp16)[name = tensor("linear_240_cast_fp16")]; + tensor input_597_cast_fp16 = add(x = linear_240_cast_fp16, y = hidden_states_497_cast_fp16)[name = tensor("input_597_cast_fp16")]; + tensor input_599_axes_0 = const()[name = tensor("input_599_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(575517184)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(575519808)))]; + tensor input_599_cast_fp16 = layer_norm(axes = input_599_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_8_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_8_norm2_weight_to_fp16, x = input_597_cast_fp16)[name = tensor("input_599_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(575522432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(576751296))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_241_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized, x = input_599_cast_fp16)[name = tensor("linear_241_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(576751488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(578717632))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_242_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_242_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(578717824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(580683968))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_243_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_243_cast_fp16")]; + tensor var_3462 = const()[name = tensor("op_3462"), val = tensor([1, -1, 20, 64])]; + tensor var_3463_cast_fp16 = reshape(shape = var_3462, x = linear_241_cast_fp16)[name = tensor("op_3463_cast_fp16")]; + tensor var_3465 = const()[name = tensor("op_3465"), val = tensor([1, -1, 20, 64])]; + tensor var_3466_cast_fp16 = reshape(shape = var_3465, x = linear_242_cast_fp16)[name = tensor("op_3466_cast_fp16")]; + tensor var_3468 = const()[name = tensor("op_3468"), val = tensor([1, -1, 20, 64])]; + tensor var_3469_cast_fp16 = reshape(shape = var_3468, x = linear_243_cast_fp16)[name = tensor("op_3469_cast_fp16")]; + tensor value_183_perm_0 = const()[name = tensor("value_183_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_61_y_0_to_fp16 = const()[name = tensor("mul_61_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_61_cast_fp16 = mul(x = var_3463_cast_fp16, y = mul_61_y_0_to_fp16)[name = tensor("mul_61_cast_fp16")]; + tensor matmul_45_transpose_y_0 = const()[name = tensor("matmul_45_transpose_y_0"), val = tensor(true)]; + tensor matmul_45_transpose_x_0 = const()[name = tensor("matmul_45_transpose_x_0"), val = tensor(false)]; + tensor transpose_362_perm_0 = const()[name = tensor("transpose_362_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_363_perm_0 = const()[name = tensor("transpose_363_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_501 = transpose(perm = transpose_363_perm_0, x = var_3466_cast_fp16)[name = tensor("transpose_501")]; + tensor transpose_502 = transpose(perm = transpose_362_perm_0, x = mul_61_cast_fp16)[name = tensor("transpose_502")]; + tensor matmul_45_cast_fp16 = matmul(transpose_x = matmul_45_transpose_x_0, transpose_y = matmul_45_transpose_y_0, x = transpose_502, y = transpose_501)[name = tensor("matmul_45_cast_fp16")]; + tensor softmax_45_axis_0 = const()[name = tensor("softmax_45_axis_0"), val = tensor(-1)]; + tensor softmax_45_cast_fp16 = softmax(axis = softmax_45_axis_0, x = matmul_45_cast_fp16)[name = tensor("softmax_45_cast_fp16")]; + tensor hidden_states_507_transpose_x_0 = const()[name = tensor("hidden_states_507_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_507_transpose_y_0 = const()[name = tensor("hidden_states_507_transpose_y_0"), val = tensor(false)]; + tensor transpose_500 = transpose(perm = value_183_perm_0, x = var_3469_cast_fp16)[name = tensor("transpose_500")]; + tensor hidden_states_507_cast_fp16 = matmul(transpose_x = hidden_states_507_transpose_x_0, transpose_y = hidden_states_507_transpose_y_0, x = softmax_45_cast_fp16, y = transpose_500)[name = tensor("hidden_states_507_cast_fp16")]; + tensor var_3472_perm_0 = const()[name = tensor("op_3472_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3476 = const()[name = tensor("op_3476"), val = tensor([1, -1, 1280])]; + tensor transpose_499 = transpose(perm = var_3472_perm_0, x = hidden_states_507_cast_fp16)[name = tensor("transpose_499")]; + tensor hidden_states_509_cast_fp16 = reshape(shape = var_3476, x = transpose_499)[name = tensor("hidden_states_509_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(580684160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(581913024))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(581913216)))]; + tensor linear_244_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_509_cast_fp16)[name = tensor("linear_244_cast_fp16")]; + tensor input_605_cast_fp16 = add(x = linear_244_cast_fp16, y = input_597_cast_fp16)[name = tensor("input_605_cast_fp16")]; + tensor input_607_axes_0 = const()[name = tensor("input_607_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(581915840)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(581918464)))]; + tensor input_607_cast_fp16 = layer_norm(axes = input_607_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_8_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_8_norm3_weight_to_fp16, x = input_605_cast_fp16)[name = tensor("input_607_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(581921088))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(591751552))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(591751744))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(591759488))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_245_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized, x = input_607_cast_fp16)[name = tensor("linear_245_cast_fp16")]; + tensor var_3498_split_sizes_0 = const()[name = tensor("op_3498_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3498_axis_0 = const()[name = tensor("op_3498_axis_0"), val = tensor(-1)]; + tensor var_3498_cast_fp16_0, tensor var_3498_cast_fp16_1 = split(axis = var_3498_axis_0, split_sizes = var_3498_split_sizes_0, x = linear_245_cast_fp16)[name = tensor("op_3498_cast_fp16")]; + tensor var_3500_mode_0 = const()[name = tensor("op_3500_mode_0"), val = tensor("EXACT")]; + tensor var_3500_cast_fp16 = gelu(mode = var_3500_mode_0, x = var_3498_cast_fp16_1)[name = tensor("op_3500_cast_fp16")]; + tensor input_609_cast_fp16 = mul(x = var_3498_cast_fp16_0, y = var_3500_cast_fp16)[name = tensor("input_609_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(591759680))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(596674944))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(596675136)))]; + tensor linear_246_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized, x = input_609_cast_fp16)[name = tensor("linear_246_cast_fp16")]; + tensor hidden_states_517_cast_fp16 = add(x = linear_246_cast_fp16, y = input_605_cast_fp16)[name = tensor("hidden_states_517_cast_fp16")]; + tensor hidden_states_519_axes_0 = const()[name = tensor("hidden_states_519_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm1_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(596677760)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm1_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(596680384)))]; + tensor hidden_states_519_cast_fp16 = layer_norm(axes = hidden_states_519_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_9_norm1_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_9_norm1_weight_to_fp16, x = hidden_states_517_cast_fp16)[name = tensor("hidden_states_519_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(596683008))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(597911872))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_247_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_519_cast_fp16)[name = tensor("linear_247_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(597912064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(599140928))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_248_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_519_cast_fp16)[name = tensor("linear_248_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(599141120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(600369984))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_249_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_519_cast_fp16)[name = tensor("linear_249_cast_fp16")]; + tensor var_3535 = const()[name = tensor("op_3535"), val = tensor([1, -1, 20, 64])]; + tensor var_3536_cast_fp16 = reshape(shape = var_3535, x = linear_247_cast_fp16)[name = tensor("op_3536_cast_fp16")]; + tensor var_3538 = const()[name = tensor("op_3538"), val = tensor([1, -1, 20, 64])]; + tensor var_3539_cast_fp16 = reshape(shape = var_3538, x = linear_248_cast_fp16)[name = tensor("op_3539_cast_fp16")]; + tensor var_3541 = const()[name = tensor("op_3541"), val = tensor([1, -1, 20, 64])]; + tensor var_3542_cast_fp16 = reshape(shape = var_3541, x = linear_249_cast_fp16)[name = tensor("op_3542_cast_fp16")]; + tensor value_187_perm_0 = const()[name = tensor("value_187_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_62_y_0_to_fp16 = const()[name = tensor("mul_62_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_62_cast_fp16 = mul(x = var_3536_cast_fp16, y = mul_62_y_0_to_fp16)[name = tensor("mul_62_cast_fp16")]; + tensor matmul_46_transpose_y_0 = const()[name = tensor("matmul_46_transpose_y_0"), val = tensor(true)]; + tensor matmul_46_transpose_x_0 = const()[name = tensor("matmul_46_transpose_x_0"), val = tensor(false)]; + tensor transpose_364_perm_0 = const()[name = tensor("transpose_364_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_365_perm_0 = const()[name = tensor("transpose_365_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_497 = transpose(perm = transpose_365_perm_0, x = var_3539_cast_fp16)[name = tensor("transpose_497")]; + tensor transpose_498 = transpose(perm = transpose_364_perm_0, x = mul_62_cast_fp16)[name = tensor("transpose_498")]; + tensor matmul_46_cast_fp16 = matmul(transpose_x = matmul_46_transpose_x_0, transpose_y = matmul_46_transpose_y_0, x = transpose_498, y = transpose_497)[name = tensor("matmul_46_cast_fp16")]; + tensor softmax_46_axis_0 = const()[name = tensor("softmax_46_axis_0"), val = tensor(-1)]; + tensor softmax_46_cast_fp16 = softmax(axis = softmax_46_axis_0, x = matmul_46_cast_fp16)[name = tensor("softmax_46_cast_fp16")]; + tensor hidden_states_521_transpose_x_0 = const()[name = tensor("hidden_states_521_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_521_transpose_y_0 = const()[name = tensor("hidden_states_521_transpose_y_0"), val = tensor(false)]; + tensor transpose_496 = transpose(perm = value_187_perm_0, x = var_3542_cast_fp16)[name = tensor("transpose_496")]; + tensor hidden_states_521_cast_fp16 = matmul(transpose_x = hidden_states_521_transpose_x_0, transpose_y = hidden_states_521_transpose_y_0, x = softmax_46_cast_fp16, y = transpose_496)[name = tensor("hidden_states_521_cast_fp16")]; + tensor var_3545_perm_0 = const()[name = tensor("op_3545_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3549 = const()[name = tensor("op_3549"), val = tensor([1, -1, 1280])]; + tensor transpose_495 = transpose(perm = var_3545_perm_0, x = hidden_states_521_cast_fp16)[name = tensor("transpose_495")]; + tensor hidden_states_523_cast_fp16 = reshape(shape = var_3549, x = transpose_495)[name = tensor("hidden_states_523_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(600370176))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(601599040))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(601599232)))]; + tensor linear_250_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_523_cast_fp16)[name = tensor("linear_250_cast_fp16")]; + tensor input_617_cast_fp16 = add(x = linear_250_cast_fp16, y = hidden_states_517_cast_fp16)[name = tensor("input_617_cast_fp16")]; + tensor input_619_axes_0 = const()[name = tensor("input_619_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm2_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(601601856)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(601604480)))]; + tensor input_619_cast_fp16 = layer_norm(axes = input_619_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_9_norm2_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_9_norm2_weight_to_fp16, x = input_617_cast_fp16)[name = tensor("input_619_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(601607104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(602835968))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_251_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized, x = input_619_cast_fp16)[name = tensor("linear_251_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(602836160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(604802304))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_252_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_252_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(604802496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(606768640))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_253_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_253_cast_fp16")]; + tensor var_3581 = const()[name = tensor("op_3581"), val = tensor([1, -1, 20, 64])]; + tensor var_3582_cast_fp16 = reshape(shape = var_3581, x = linear_251_cast_fp16)[name = tensor("op_3582_cast_fp16")]; + tensor var_3584 = const()[name = tensor("op_3584"), val = tensor([1, -1, 20, 64])]; + tensor var_3585_cast_fp16 = reshape(shape = var_3584, x = linear_252_cast_fp16)[name = tensor("op_3585_cast_fp16")]; + tensor var_3587 = const()[name = tensor("op_3587"), val = tensor([1, -1, 20, 64])]; + tensor var_3588_cast_fp16 = reshape(shape = var_3587, x = linear_253_cast_fp16)[name = tensor("op_3588_cast_fp16")]; + tensor value_191_perm_0 = const()[name = tensor("value_191_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_63_y_0_to_fp16 = const()[name = tensor("mul_63_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_63_cast_fp16 = mul(x = var_3582_cast_fp16, y = mul_63_y_0_to_fp16)[name = tensor("mul_63_cast_fp16")]; + tensor matmul_47_transpose_y_0 = const()[name = tensor("matmul_47_transpose_y_0"), val = tensor(true)]; + tensor matmul_47_transpose_x_0 = const()[name = tensor("matmul_47_transpose_x_0"), val = tensor(false)]; + tensor transpose_366_perm_0 = const()[name = tensor("transpose_366_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_367_perm_0 = const()[name = tensor("transpose_367_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_493 = transpose(perm = transpose_367_perm_0, x = var_3585_cast_fp16)[name = tensor("transpose_493")]; + tensor transpose_494 = transpose(perm = transpose_366_perm_0, x = mul_63_cast_fp16)[name = tensor("transpose_494")]; + tensor matmul_47_cast_fp16 = matmul(transpose_x = matmul_47_transpose_x_0, transpose_y = matmul_47_transpose_y_0, x = transpose_494, y = transpose_493)[name = tensor("matmul_47_cast_fp16")]; + tensor softmax_47_axis_0 = const()[name = tensor("softmax_47_axis_0"), val = tensor(-1)]; + tensor softmax_47_cast_fp16 = softmax(axis = softmax_47_axis_0, x = matmul_47_cast_fp16)[name = tensor("softmax_47_cast_fp16")]; + tensor hidden_states_527_transpose_x_0 = const()[name = tensor("hidden_states_527_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_527_transpose_y_0 = const()[name = tensor("hidden_states_527_transpose_y_0"), val = tensor(false)]; + tensor transpose_492 = transpose(perm = value_191_perm_0, x = var_3588_cast_fp16)[name = tensor("transpose_492")]; + tensor hidden_states_527_cast_fp16 = matmul(transpose_x = hidden_states_527_transpose_x_0, transpose_y = hidden_states_527_transpose_y_0, x = softmax_47_cast_fp16, y = transpose_492)[name = tensor("hidden_states_527_cast_fp16")]; + tensor var_3591_perm_0 = const()[name = tensor("op_3591_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3595 = const()[name = tensor("op_3595"), val = tensor([1, -1, 1280])]; + tensor transpose_491 = transpose(perm = var_3591_perm_0, x = hidden_states_527_cast_fp16)[name = tensor("transpose_491")]; + tensor hidden_states_529_cast_fp16 = reshape(shape = var_3595, x = transpose_491)[name = tensor("hidden_states_529_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(606768832))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(607997696))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(607997888)))]; + tensor linear_254_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_529_cast_fp16)[name = tensor("linear_254_cast_fp16")]; + tensor input_625_cast_fp16 = add(x = linear_254_cast_fp16, y = input_617_cast_fp16)[name = tensor("input_625_cast_fp16")]; + tensor input_627_axes_0 = const()[name = tensor("input_627_axes_0"), val = tensor([-1])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm3_weight_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(608000512)))]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_norm3_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(608003136)))]; + tensor input_627_cast_fp16 = layer_norm(axes = input_627_axes_0, beta = down_blocks_2_attentions_1_transformer_blocks_9_norm3_bias_to_fp16, epsilon = var_1044_to_fp16, gamma = down_blocks_2_attentions_1_transformer_blocks_9_norm3_weight_to_fp16, x = input_625_cast_fp16)[name = tensor("input_627_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(608005760))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(617836224))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(617836416))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(617844160))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_255_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized, weight = down_blocks_2_attentions_1_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized, x = input_627_cast_fp16)[name = tensor("linear_255_cast_fp16")]; + tensor var_3617_split_sizes_0 = const()[name = tensor("op_3617_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3617_axis_0 = const()[name = tensor("op_3617_axis_0"), val = tensor(-1)]; + tensor var_3617_cast_fp16_0, tensor var_3617_cast_fp16_1 = split(axis = var_3617_axis_0, split_sizes = var_3617_split_sizes_0, x = linear_255_cast_fp16)[name = tensor("op_3617_cast_fp16")]; + tensor var_3619_mode_0 = const()[name = tensor("op_3619_mode_0"), val = tensor("EXACT")]; + tensor var_3619_cast_fp16 = gelu(mode = var_3619_mode_0, x = var_3617_cast_fp16_1)[name = tensor("op_3619_cast_fp16")]; + tensor input_629_cast_fp16 = mul(x = var_3617_cast_fp16_0, y = var_3619_cast_fp16)[name = tensor("input_629_cast_fp16")]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(617844352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(622759616))), name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(622759808)))]; + tensor linear_256_cast_fp16 = linear(bias = down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_bias_to_fp16, weight = down_blocks_2_attentions_1_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized, x = input_629_cast_fp16)[name = tensor("linear_256_cast_fp16")]; + tensor input_633_cast_fp16 = add(x = linear_256_cast_fp16, y = input_625_cast_fp16)[name = tensor("input_633_cast_fp16")]; + tensor down_blocks_2_attentions_1_proj_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(622762432))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(623991296))), name = tensor("down_blocks_2_attentions_1_proj_out_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor down_blocks_2_attentions_1_proj_out_bias_to_fp16 = const()[name = tensor("down_blocks_2_attentions_1_proj_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(623991488)))]; + tensor linear_257_cast_fp16 = linear(bias = down_blocks_2_attentions_1_proj_out_bias_to_fp16, weight = down_blocks_2_attentions_1_proj_out_weight_to_fp16_palettized, x = input_633_cast_fp16)[name = tensor("linear_257_cast_fp16")]; + tensor var_3629 = const()[name = tensor("op_3629"), val = tensor([1, 32, 32, 1280])]; + tensor var_3630_cast_fp16 = reshape(shape = var_3629, x = linear_257_cast_fp16)[name = tensor("op_3630_cast_fp16")]; + tensor var_3631 = const()[name = tensor("op_3631"), val = tensor([0, 3, 1, 2])]; + tensor transpose_490 = transpose(perm = var_3631, x = var_3630_cast_fp16)[name = tensor("transpose_490")]; + tensor input_635_cast_fp16 = add(x = transpose_490, y = var_2389_cast_fp16)[name = tensor("input_635_cast_fp16")]; + tensor var_3651 = const()[name = tensor("op_3651"), val = tensor(1)]; + tensor reshape_64_shape_0 = const()[name = tensor("reshape_64_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_64_cast_fp16 = reshape(shape = reshape_64_shape_0, x = input_635_cast_fp16)[name = tensor("reshape_64_cast_fp16")]; + tensor reduce_mean_48_axes_0 = const()[name = tensor("reduce_mean_48_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_48_keep_dims_0 = const()[name = tensor("reduce_mean_48_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_48_cast_fp16 = reduce_mean(axes = reduce_mean_48_axes_0, keep_dims = reduce_mean_48_keep_dims_0, x = reshape_64_cast_fp16)[name = tensor("reduce_mean_48_cast_fp16")]; + tensor sub_32_cast_fp16 = sub(x = reshape_64_cast_fp16, y = reduce_mean_48_cast_fp16)[name = tensor("sub_32_cast_fp16")]; + tensor square_16_cast_fp16 = square(x = sub_32_cast_fp16)[name = tensor("square_16_cast_fp16")]; + tensor reduce_mean_50_axes_0 = const()[name = tensor("reduce_mean_50_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_50_keep_dims_0 = const()[name = tensor("reduce_mean_50_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_50_cast_fp16 = reduce_mean(axes = reduce_mean_50_axes_0, keep_dims = reduce_mean_50_keep_dims_0, x = square_16_cast_fp16)[name = tensor("reduce_mean_50_cast_fp16")]; + tensor add_32_y_0_to_fp16 = const()[name = tensor("add_32_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_32_cast_fp16 = add(x = reduce_mean_50_cast_fp16, y = add_32_y_0_to_fp16)[name = tensor("add_32_cast_fp16")]; + tensor sqrt_16_cast_fp16 = sqrt(x = add_32_cast_fp16)[name = tensor("sqrt_16_cast_fp16")]; + tensor real_div_16_cast_fp16 = real_div(x = sub_32_cast_fp16, y = sqrt_16_cast_fp16)[name = tensor("real_div_16_cast_fp16")]; + tensor reshape_65_shape_0 = const()[name = tensor("reshape_65_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_65_cast_fp16 = reshape(shape = reshape_65_shape_0, x = real_div_16_cast_fp16)[name = tensor("reshape_65_cast_fp16")]; + tensor add_33_gamma_0_to_fp16 = const()[name = tensor("add_33_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(623994112)))]; + tensor add_33_beta_0_to_fp16 = const()[name = tensor("add_33_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(623996736)))]; + tensor add_33_epsilon_0_to_fp16 = const()[name = tensor("add_33_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_33_cast_fp16 = batch_norm(beta = add_33_beta_0_to_fp16, epsilon = add_33_epsilon_0_to_fp16, gamma = add_33_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_65_cast_fp16)[name = tensor("add_33_cast_fp16")]; + tensor input_639_cast_fp16 = silu(x = add_33_cast_fp16)[name = tensor("input_639_cast_fp16")]; + tensor var_3673 = const()[name = tensor("op_3673"), val = tensor([1, 1])]; + tensor var_3675 = const()[name = tensor("op_3675"), val = tensor([1, 1])]; + tensor hidden_states_541_pad_type_0 = const()[name = tensor("hidden_states_541_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_541_pad_0 = const()[name = tensor("hidden_states_541_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor mid_block_resnets_0_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(623999360))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(635058624))), name = tensor("mid_block_resnets_0_conv1_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor mid_block_resnets_0_conv1_bias_to_fp16 = const()[name = tensor("mid_block_resnets_0_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(635058816)))]; + tensor hidden_states_541_cast_fp16 = conv(bias = mid_block_resnets_0_conv1_bias_to_fp16, dilations = var_3675, groups = var_3651, pad = hidden_states_541_pad_0, pad_type = hidden_states_541_pad_type_0, strides = var_3673, weight = mid_block_resnets_0_conv1_weight_to_fp16_palettized, x = input_639_cast_fp16)[name = tensor("hidden_states_541_cast_fp16")]; + tensor mid_block_resnets_0_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(635061440))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(636290304))), name = tensor("mid_block_resnets_0_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_resnets_0_time_emb_proj_bias_to_fp16 = const()[name = tensor("mid_block_resnets_0_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(636290496)))]; + tensor linear_258_cast_fp16 = linear(bias = mid_block_resnets_0_time_emb_proj_bias_to_fp16, weight = mid_block_resnets_0_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_258_cast_fp16")]; + tensor var_3684_axes_0 = const()[name = tensor("op_3684_axes_0"), val = tensor([2])]; + tensor var_3684_cast_fp16 = expand_dims(axes = var_3684_axes_0, x = linear_258_cast_fp16)[name = tensor("op_3684_cast_fp16")]; + tensor temb_13_axes_0 = const()[name = tensor("temb_13_axes_0"), val = tensor([3])]; + tensor temb_13_cast_fp16 = expand_dims(axes = temb_13_axes_0, x = var_3684_cast_fp16)[name = tensor("temb_13_cast_fp16")]; + tensor input_643_cast_fp16 = add(x = hidden_states_541_cast_fp16, y = temb_13_cast_fp16)[name = tensor("input_643_cast_fp16")]; + tensor reshape_68_shape_0 = const()[name = tensor("reshape_68_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_68_cast_fp16 = reshape(shape = reshape_68_shape_0, x = input_643_cast_fp16)[name = tensor("reshape_68_cast_fp16")]; + tensor reduce_mean_51_axes_0 = const()[name = tensor("reduce_mean_51_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_51_keep_dims_0 = const()[name = tensor("reduce_mean_51_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_51_cast_fp16 = reduce_mean(axes = reduce_mean_51_axes_0, keep_dims = reduce_mean_51_keep_dims_0, x = reshape_68_cast_fp16)[name = tensor("reduce_mean_51_cast_fp16")]; + tensor sub_34_cast_fp16 = sub(x = reshape_68_cast_fp16, y = reduce_mean_51_cast_fp16)[name = tensor("sub_34_cast_fp16")]; + tensor square_17_cast_fp16 = square(x = sub_34_cast_fp16)[name = tensor("square_17_cast_fp16")]; + tensor reduce_mean_53_axes_0 = const()[name = tensor("reduce_mean_53_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_53_keep_dims_0 = const()[name = tensor("reduce_mean_53_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_53_cast_fp16 = reduce_mean(axes = reduce_mean_53_axes_0, keep_dims = reduce_mean_53_keep_dims_0, x = square_17_cast_fp16)[name = tensor("reduce_mean_53_cast_fp16")]; + tensor add_34_y_0_to_fp16 = const()[name = tensor("add_34_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_34_cast_fp16 = add(x = reduce_mean_53_cast_fp16, y = add_34_y_0_to_fp16)[name = tensor("add_34_cast_fp16")]; + tensor sqrt_17_cast_fp16 = sqrt(x = add_34_cast_fp16)[name = tensor("sqrt_17_cast_fp16")]; + tensor real_div_17_cast_fp16 = real_div(x = sub_34_cast_fp16, y = sqrt_17_cast_fp16)[name = tensor("real_div_17_cast_fp16")]; + tensor reshape_69_shape_0 = const()[name = tensor("reshape_69_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_69_cast_fp16 = reshape(shape = reshape_69_shape_0, x = real_div_17_cast_fp16)[name = tensor("reshape_69_cast_fp16")]; + tensor add_35_gamma_0_to_fp16 = const()[name = tensor("add_35_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(636293120)))]; + tensor add_35_beta_0_to_fp16 = const()[name = tensor("add_35_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(636295744)))]; + tensor add_35_epsilon_0_to_fp16 = const()[name = tensor("add_35_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_35_cast_fp16 = batch_norm(beta = add_35_beta_0_to_fp16, epsilon = add_35_epsilon_0_to_fp16, gamma = add_35_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_69_cast_fp16)[name = tensor("add_35_cast_fp16")]; + tensor input_647_cast_fp16 = silu(x = add_35_cast_fp16)[name = tensor("input_647_cast_fp16")]; + tensor var_3694 = const()[name = tensor("op_3694"), val = tensor([1, 1])]; + tensor var_3696 = const()[name = tensor("op_3696"), val = tensor([1, 1])]; + tensor hidden_states_543_pad_type_0 = const()[name = tensor("hidden_states_543_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_543_pad_0 = const()[name = tensor("hidden_states_543_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor mid_block_resnets_0_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(636298368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(647357632))), name = tensor("mid_block_resnets_0_conv2_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor mid_block_resnets_0_conv2_bias_to_fp16 = const()[name = tensor("mid_block_resnets_0_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(647357824)))]; + tensor hidden_states_543_cast_fp16 = conv(bias = mid_block_resnets_0_conv2_bias_to_fp16, dilations = var_3696, groups = var_3651, pad = hidden_states_543_pad_0, pad_type = hidden_states_543_pad_type_0, strides = var_3694, weight = mid_block_resnets_0_conv2_weight_to_fp16_palettized, x = input_647_cast_fp16)[name = tensor("hidden_states_543_cast_fp16")]; + tensor var_3699_cast_fp16 = add(x = input_635_cast_fp16, y = hidden_states_543_cast_fp16)[name = tensor("op_3699_cast_fp16")]; + tensor reshape_72_shape_0 = const()[name = tensor("reshape_72_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_72_cast_fp16 = reshape(shape = reshape_72_shape_0, x = var_3699_cast_fp16)[name = tensor("reshape_72_cast_fp16")]; + tensor reduce_mean_54_axes_0 = const()[name = tensor("reduce_mean_54_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_54_keep_dims_0 = const()[name = tensor("reduce_mean_54_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_54_cast_fp16 = reduce_mean(axes = reduce_mean_54_axes_0, keep_dims = reduce_mean_54_keep_dims_0, x = reshape_72_cast_fp16)[name = tensor("reduce_mean_54_cast_fp16")]; + tensor sub_36_cast_fp16 = sub(x = reshape_72_cast_fp16, y = reduce_mean_54_cast_fp16)[name = tensor("sub_36_cast_fp16")]; + tensor square_18_cast_fp16 = square(x = sub_36_cast_fp16)[name = tensor("square_18_cast_fp16")]; + tensor reduce_mean_56_axes_0 = const()[name = tensor("reduce_mean_56_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_56_keep_dims_0 = const()[name = tensor("reduce_mean_56_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_56_cast_fp16 = reduce_mean(axes = reduce_mean_56_axes_0, keep_dims = reduce_mean_56_keep_dims_0, x = square_18_cast_fp16)[name = tensor("reduce_mean_56_cast_fp16")]; + tensor add_36_y_0_to_fp16 = const()[name = tensor("add_36_y_0_to_fp16"), val = tensor(0x1.1p-20)]; + tensor add_36_cast_fp16 = add(x = reduce_mean_56_cast_fp16, y = add_36_y_0_to_fp16)[name = tensor("add_36_cast_fp16")]; + tensor sqrt_18_cast_fp16 = sqrt(x = add_36_cast_fp16)[name = tensor("sqrt_18_cast_fp16")]; + tensor real_div_18_cast_fp16 = real_div(x = sub_36_cast_fp16, y = sqrt_18_cast_fp16)[name = tensor("real_div_18_cast_fp16")]; + tensor reshape_73_shape_0 = const()[name = tensor("reshape_73_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_73_cast_fp16 = reshape(shape = reshape_73_shape_0, x = real_div_18_cast_fp16)[name = tensor("reshape_73_cast_fp16")]; + tensor add_37_gamma_0_to_fp16 = const()[name = tensor("add_37_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(647360448)))]; + tensor add_37_beta_0_to_fp16 = const()[name = tensor("add_37_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(647363072)))]; + tensor add_37_epsilon_0_to_fp16 = const()[name = tensor("add_37_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_37_cast_fp16 = batch_norm(beta = add_37_beta_0_to_fp16, epsilon = add_37_epsilon_0_to_fp16, gamma = add_37_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_73_cast_fp16)[name = tensor("add_37_cast_fp16")]; + tensor var_3737 = const()[name = tensor("op_3737"), val = tensor([0, 2, 3, 1])]; + tensor var_3741 = const()[name = tensor("op_3741"), val = tensor([1, 1024, 1280])]; + tensor transpose_489 = transpose(perm = var_3737, x = add_37_cast_fp16)[name = tensor("transpose_489")]; + tensor input_651_cast_fp16 = reshape(shape = var_3741, x = transpose_489)[name = tensor("input_651_cast_fp16")]; + tensor mid_block_attentions_0_proj_in_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(647365696))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(648594560))), name = tensor("mid_block_attentions_0_proj_in_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_proj_in_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_proj_in_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(648594752)))]; + tensor linear_259_cast_fp16 = linear(bias = mid_block_attentions_0_proj_in_bias_to_fp16, weight = mid_block_attentions_0_proj_in_weight_to_fp16_palettized, x = input_651_cast_fp16)[name = tensor("linear_259_cast_fp16")]; + tensor hidden_states_551_axes_0 = const()[name = tensor("hidden_states_551_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_0_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(648597376)))]; + tensor mid_block_attentions_0_transformer_blocks_0_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(648600000)))]; + tensor var_3647_to_fp16 = const()[name = tensor("op_3647_to_fp16"), val = tensor(0x1.5p-17)]; + tensor hidden_states_551_cast_fp16 = layer_norm(axes = hidden_states_551_axes_0, beta = mid_block_attentions_0_transformer_blocks_0_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_0_norm1_weight_to_fp16, x = linear_259_cast_fp16)[name = tensor("hidden_states_551_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(648602624))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(649831488))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_260_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_551_cast_fp16)[name = tensor("linear_260_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(649831680))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(651060544))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_261_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_551_cast_fp16)[name = tensor("linear_261_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(651060736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(652289600))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_262_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_551_cast_fp16)[name = tensor("linear_262_cast_fp16")]; + tensor var_3774 = const()[name = tensor("op_3774"), val = tensor([1, -1, 20, 64])]; + tensor var_3775_cast_fp16 = reshape(shape = var_3774, x = linear_260_cast_fp16)[name = tensor("op_3775_cast_fp16")]; + tensor var_3777 = const()[name = tensor("op_3777"), val = tensor([1, -1, 20, 64])]; + tensor var_3778_cast_fp16 = reshape(shape = var_3777, x = linear_261_cast_fp16)[name = tensor("op_3778_cast_fp16")]; + tensor var_3780 = const()[name = tensor("op_3780"), val = tensor([1, -1, 20, 64])]; + tensor var_3781_cast_fp16 = reshape(shape = var_3780, x = linear_262_cast_fp16)[name = tensor("op_3781_cast_fp16")]; + tensor value_195_perm_0 = const()[name = tensor("value_195_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_67_y_0_to_fp16 = const()[name = tensor("mul_67_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_67_cast_fp16 = mul(x = var_3775_cast_fp16, y = mul_67_y_0_to_fp16)[name = tensor("mul_67_cast_fp16")]; + tensor matmul_48_transpose_y_0 = const()[name = tensor("matmul_48_transpose_y_0"), val = tensor(true)]; + tensor matmul_48_transpose_x_0 = const()[name = tensor("matmul_48_transpose_x_0"), val = tensor(false)]; + tensor transpose_368_perm_0 = const()[name = tensor("transpose_368_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_369_perm_0 = const()[name = tensor("transpose_369_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_487 = transpose(perm = transpose_369_perm_0, x = var_3778_cast_fp16)[name = tensor("transpose_487")]; + tensor transpose_488 = transpose(perm = transpose_368_perm_0, x = mul_67_cast_fp16)[name = tensor("transpose_488")]; + tensor matmul_48_cast_fp16 = matmul(transpose_x = matmul_48_transpose_x_0, transpose_y = matmul_48_transpose_y_0, x = transpose_488, y = transpose_487)[name = tensor("matmul_48_cast_fp16")]; + tensor softmax_48_axis_0 = const()[name = tensor("softmax_48_axis_0"), val = tensor(-1)]; + tensor softmax_48_cast_fp16 = softmax(axis = softmax_48_axis_0, x = matmul_48_cast_fp16)[name = tensor("softmax_48_cast_fp16")]; + tensor hidden_states_553_transpose_x_0 = const()[name = tensor("hidden_states_553_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_553_transpose_y_0 = const()[name = tensor("hidden_states_553_transpose_y_0"), val = tensor(false)]; + tensor transpose_486 = transpose(perm = value_195_perm_0, x = var_3781_cast_fp16)[name = tensor("transpose_486")]; + tensor hidden_states_553_cast_fp16 = matmul(transpose_x = hidden_states_553_transpose_x_0, transpose_y = hidden_states_553_transpose_y_0, x = softmax_48_cast_fp16, y = transpose_486)[name = tensor("hidden_states_553_cast_fp16")]; + tensor var_3784_perm_0 = const()[name = tensor("op_3784_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3788 = const()[name = tensor("op_3788"), val = tensor([1, -1, 1280])]; + tensor transpose_485 = transpose(perm = var_3784_perm_0, x = hidden_states_553_cast_fp16)[name = tensor("transpose_485")]; + tensor hidden_states_555_cast_fp16 = reshape(shape = var_3788, x = transpose_485)[name = tensor("hidden_states_555_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(652289792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(653518656))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(653518848)))]; + tensor linear_263_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_555_cast_fp16)[name = tensor("linear_263_cast_fp16")]; + tensor input_657_cast_fp16 = add(x = linear_263_cast_fp16, y = linear_259_cast_fp16)[name = tensor("input_657_cast_fp16")]; + tensor input_659_axes_0 = const()[name = tensor("input_659_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_0_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(653521472)))]; + tensor mid_block_attentions_0_transformer_blocks_0_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(653524096)))]; + tensor input_659_cast_fp16 = layer_norm(axes = input_659_axes_0, beta = mid_block_attentions_0_transformer_blocks_0_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_0_norm2_weight_to_fp16, x = input_657_cast_fp16)[name = tensor("input_659_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(653526720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(654755584))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_264_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn2_to_q_weight_to_fp16_palettized, x = input_659_cast_fp16)[name = tensor("linear_264_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(654755776))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(656721920))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_265_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_265_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(656722112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(658688256))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_266_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_266_cast_fp16")]; + tensor var_3820 = const()[name = tensor("op_3820"), val = tensor([1, -1, 20, 64])]; + tensor var_3821_cast_fp16 = reshape(shape = var_3820, x = linear_264_cast_fp16)[name = tensor("op_3821_cast_fp16")]; + tensor var_3823 = const()[name = tensor("op_3823"), val = tensor([1, -1, 20, 64])]; + tensor var_3824_cast_fp16 = reshape(shape = var_3823, x = linear_265_cast_fp16)[name = tensor("op_3824_cast_fp16")]; + tensor var_3826 = const()[name = tensor("op_3826"), val = tensor([1, -1, 20, 64])]; + tensor var_3827_cast_fp16 = reshape(shape = var_3826, x = linear_266_cast_fp16)[name = tensor("op_3827_cast_fp16")]; + tensor value_199_perm_0 = const()[name = tensor("value_199_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_68_y_0_to_fp16 = const()[name = tensor("mul_68_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_68_cast_fp16 = mul(x = var_3821_cast_fp16, y = mul_68_y_0_to_fp16)[name = tensor("mul_68_cast_fp16")]; + tensor matmul_49_transpose_y_0 = const()[name = tensor("matmul_49_transpose_y_0"), val = tensor(true)]; + tensor matmul_49_transpose_x_0 = const()[name = tensor("matmul_49_transpose_x_0"), val = tensor(false)]; + tensor transpose_370_perm_0 = const()[name = tensor("transpose_370_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_371_perm_0 = const()[name = tensor("transpose_371_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_483 = transpose(perm = transpose_371_perm_0, x = var_3824_cast_fp16)[name = tensor("transpose_483")]; + tensor transpose_484 = transpose(perm = transpose_370_perm_0, x = mul_68_cast_fp16)[name = tensor("transpose_484")]; + tensor matmul_49_cast_fp16 = matmul(transpose_x = matmul_49_transpose_x_0, transpose_y = matmul_49_transpose_y_0, x = transpose_484, y = transpose_483)[name = tensor("matmul_49_cast_fp16")]; + tensor softmax_49_axis_0 = const()[name = tensor("softmax_49_axis_0"), val = tensor(-1)]; + tensor softmax_49_cast_fp16 = softmax(axis = softmax_49_axis_0, x = matmul_49_cast_fp16)[name = tensor("softmax_49_cast_fp16")]; + tensor hidden_states_559_transpose_x_0 = const()[name = tensor("hidden_states_559_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_559_transpose_y_0 = const()[name = tensor("hidden_states_559_transpose_y_0"), val = tensor(false)]; + tensor transpose_482 = transpose(perm = value_199_perm_0, x = var_3827_cast_fp16)[name = tensor("transpose_482")]; + tensor hidden_states_559_cast_fp16 = matmul(transpose_x = hidden_states_559_transpose_x_0, transpose_y = hidden_states_559_transpose_y_0, x = softmax_49_cast_fp16, y = transpose_482)[name = tensor("hidden_states_559_cast_fp16")]; + tensor var_3830_perm_0 = const()[name = tensor("op_3830_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3834 = const()[name = tensor("op_3834"), val = tensor([1, -1, 1280])]; + tensor transpose_481 = transpose(perm = var_3830_perm_0, x = hidden_states_559_cast_fp16)[name = tensor("transpose_481")]; + tensor hidden_states_561_cast_fp16 = reshape(shape = var_3834, x = transpose_481)[name = tensor("hidden_states_561_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(658688448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(659917312))), name = tensor("mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(659917504)))]; + tensor linear_267_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_561_cast_fp16)[name = tensor("linear_267_cast_fp16")]; + tensor input_665_cast_fp16 = add(x = linear_267_cast_fp16, y = input_657_cast_fp16)[name = tensor("input_665_cast_fp16")]; + tensor input_667_axes_0 = const()[name = tensor("input_667_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_0_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(659920128)))]; + tensor mid_block_attentions_0_transformer_blocks_0_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(659922752)))]; + tensor input_667_cast_fp16 = layer_norm(axes = input_667_axes_0, beta = mid_block_attentions_0_transformer_blocks_0_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_0_norm3_weight_to_fp16, x = input_665_cast_fp16)[name = tensor("input_667_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(659925376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(669755840))), name = tensor("mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(669756032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(669763776))), name = tensor("mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_268_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_0_ff_net_0_proj_weight_to_fp16_palettized, x = input_667_cast_fp16)[name = tensor("linear_268_cast_fp16")]; + tensor var_3856_split_sizes_0 = const()[name = tensor("op_3856_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3856_axis_0 = const()[name = tensor("op_3856_axis_0"), val = tensor(-1)]; + tensor var_3856_cast_fp16_0, tensor var_3856_cast_fp16_1 = split(axis = var_3856_axis_0, split_sizes = var_3856_split_sizes_0, x = linear_268_cast_fp16)[name = tensor("op_3856_cast_fp16")]; + tensor var_3858_mode_0 = const()[name = tensor("op_3858_mode_0"), val = tensor("EXACT")]; + tensor var_3858_cast_fp16 = gelu(mode = var_3858_mode_0, x = var_3856_cast_fp16_1)[name = tensor("op_3858_cast_fp16")]; + tensor input_669_cast_fp16 = mul(x = var_3856_cast_fp16_0, y = var_3858_cast_fp16)[name = tensor("input_669_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(669763968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(674679232))), name = tensor("mid_block_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(674679424)))]; + tensor linear_269_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_0_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_0_ff_net_2_weight_to_fp16_palettized, x = input_669_cast_fp16)[name = tensor("linear_269_cast_fp16")]; + tensor hidden_states_569_cast_fp16 = add(x = linear_269_cast_fp16, y = input_665_cast_fp16)[name = tensor("hidden_states_569_cast_fp16")]; + tensor hidden_states_571_axes_0 = const()[name = tensor("hidden_states_571_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_1_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(674682048)))]; + tensor mid_block_attentions_0_transformer_blocks_1_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(674684672)))]; + tensor hidden_states_571_cast_fp16 = layer_norm(axes = hidden_states_571_axes_0, beta = mid_block_attentions_0_transformer_blocks_1_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_1_norm1_weight_to_fp16, x = hidden_states_569_cast_fp16)[name = tensor("hidden_states_571_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(674687296))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(675916160))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_270_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_571_cast_fp16)[name = tensor("linear_270_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(675916352))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(677145216))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_271_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_571_cast_fp16)[name = tensor("linear_271_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(677145408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(678374272))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_272_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_571_cast_fp16)[name = tensor("linear_272_cast_fp16")]; + tensor var_3893 = const()[name = tensor("op_3893"), val = tensor([1, -1, 20, 64])]; + tensor var_3894_cast_fp16 = reshape(shape = var_3893, x = linear_270_cast_fp16)[name = tensor("op_3894_cast_fp16")]; + tensor var_3896 = const()[name = tensor("op_3896"), val = tensor([1, -1, 20, 64])]; + tensor var_3897_cast_fp16 = reshape(shape = var_3896, x = linear_271_cast_fp16)[name = tensor("op_3897_cast_fp16")]; + tensor var_3899 = const()[name = tensor("op_3899"), val = tensor([1, -1, 20, 64])]; + tensor var_3900_cast_fp16 = reshape(shape = var_3899, x = linear_272_cast_fp16)[name = tensor("op_3900_cast_fp16")]; + tensor value_203_perm_0 = const()[name = tensor("value_203_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_69_y_0_to_fp16 = const()[name = tensor("mul_69_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_69_cast_fp16 = mul(x = var_3894_cast_fp16, y = mul_69_y_0_to_fp16)[name = tensor("mul_69_cast_fp16")]; + tensor matmul_50_transpose_y_0 = const()[name = tensor("matmul_50_transpose_y_0"), val = tensor(true)]; + tensor matmul_50_transpose_x_0 = const()[name = tensor("matmul_50_transpose_x_0"), val = tensor(false)]; + tensor transpose_372_perm_0 = const()[name = tensor("transpose_372_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_373_perm_0 = const()[name = tensor("transpose_373_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_479 = transpose(perm = transpose_373_perm_0, x = var_3897_cast_fp16)[name = tensor("transpose_479")]; + tensor transpose_480 = transpose(perm = transpose_372_perm_0, x = mul_69_cast_fp16)[name = tensor("transpose_480")]; + tensor matmul_50_cast_fp16 = matmul(transpose_x = matmul_50_transpose_x_0, transpose_y = matmul_50_transpose_y_0, x = transpose_480, y = transpose_479)[name = tensor("matmul_50_cast_fp16")]; + tensor softmax_50_axis_0 = const()[name = tensor("softmax_50_axis_0"), val = tensor(-1)]; + tensor softmax_50_cast_fp16 = softmax(axis = softmax_50_axis_0, x = matmul_50_cast_fp16)[name = tensor("softmax_50_cast_fp16")]; + tensor hidden_states_573_transpose_x_0 = const()[name = tensor("hidden_states_573_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_573_transpose_y_0 = const()[name = tensor("hidden_states_573_transpose_y_0"), val = tensor(false)]; + tensor transpose_478 = transpose(perm = value_203_perm_0, x = var_3900_cast_fp16)[name = tensor("transpose_478")]; + tensor hidden_states_573_cast_fp16 = matmul(transpose_x = hidden_states_573_transpose_x_0, transpose_y = hidden_states_573_transpose_y_0, x = softmax_50_cast_fp16, y = transpose_478)[name = tensor("hidden_states_573_cast_fp16")]; + tensor var_3903_perm_0 = const()[name = tensor("op_3903_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3907 = const()[name = tensor("op_3907"), val = tensor([1, -1, 1280])]; + tensor transpose_477 = transpose(perm = var_3903_perm_0, x = hidden_states_573_cast_fp16)[name = tensor("transpose_477")]; + tensor hidden_states_575_cast_fp16 = reshape(shape = var_3907, x = transpose_477)[name = tensor("hidden_states_575_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(678374464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(679603328))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(679603520)))]; + tensor linear_273_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_575_cast_fp16)[name = tensor("linear_273_cast_fp16")]; + tensor input_677_cast_fp16 = add(x = linear_273_cast_fp16, y = hidden_states_569_cast_fp16)[name = tensor("input_677_cast_fp16")]; + tensor input_679_axes_0 = const()[name = tensor("input_679_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_1_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(679606144)))]; + tensor mid_block_attentions_0_transformer_blocks_1_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(679608768)))]; + tensor input_679_cast_fp16 = layer_norm(axes = input_679_axes_0, beta = mid_block_attentions_0_transformer_blocks_1_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_1_norm2_weight_to_fp16, x = input_677_cast_fp16)[name = tensor("input_679_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(679611392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(680840256))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_274_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn2_to_q_weight_to_fp16_palettized, x = input_679_cast_fp16)[name = tensor("linear_274_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(680840448))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(682806592))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_275_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_275_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(682806784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(684772928))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_276_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_276_cast_fp16")]; + tensor var_3939 = const()[name = tensor("op_3939"), val = tensor([1, -1, 20, 64])]; + tensor var_3940_cast_fp16 = reshape(shape = var_3939, x = linear_274_cast_fp16)[name = tensor("op_3940_cast_fp16")]; + tensor var_3942 = const()[name = tensor("op_3942"), val = tensor([1, -1, 20, 64])]; + tensor var_3943_cast_fp16 = reshape(shape = var_3942, x = linear_275_cast_fp16)[name = tensor("op_3943_cast_fp16")]; + tensor var_3945 = const()[name = tensor("op_3945"), val = tensor([1, -1, 20, 64])]; + tensor var_3946_cast_fp16 = reshape(shape = var_3945, x = linear_276_cast_fp16)[name = tensor("op_3946_cast_fp16")]; + tensor value_207_perm_0 = const()[name = tensor("value_207_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_70_y_0_to_fp16 = const()[name = tensor("mul_70_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_70_cast_fp16 = mul(x = var_3940_cast_fp16, y = mul_70_y_0_to_fp16)[name = tensor("mul_70_cast_fp16")]; + tensor matmul_51_transpose_y_0 = const()[name = tensor("matmul_51_transpose_y_0"), val = tensor(true)]; + tensor matmul_51_transpose_x_0 = const()[name = tensor("matmul_51_transpose_x_0"), val = tensor(false)]; + tensor transpose_374_perm_0 = const()[name = tensor("transpose_374_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_375_perm_0 = const()[name = tensor("transpose_375_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_475 = transpose(perm = transpose_375_perm_0, x = var_3943_cast_fp16)[name = tensor("transpose_475")]; + tensor transpose_476 = transpose(perm = transpose_374_perm_0, x = mul_70_cast_fp16)[name = tensor("transpose_476")]; + tensor matmul_51_cast_fp16 = matmul(transpose_x = matmul_51_transpose_x_0, transpose_y = matmul_51_transpose_y_0, x = transpose_476, y = transpose_475)[name = tensor("matmul_51_cast_fp16")]; + tensor softmax_51_axis_0 = const()[name = tensor("softmax_51_axis_0"), val = tensor(-1)]; + tensor softmax_51_cast_fp16 = softmax(axis = softmax_51_axis_0, x = matmul_51_cast_fp16)[name = tensor("softmax_51_cast_fp16")]; + tensor hidden_states_579_transpose_x_0 = const()[name = tensor("hidden_states_579_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_579_transpose_y_0 = const()[name = tensor("hidden_states_579_transpose_y_0"), val = tensor(false)]; + tensor transpose_474 = transpose(perm = value_207_perm_0, x = var_3946_cast_fp16)[name = tensor("transpose_474")]; + tensor hidden_states_579_cast_fp16 = matmul(transpose_x = hidden_states_579_transpose_x_0, transpose_y = hidden_states_579_transpose_y_0, x = softmax_51_cast_fp16, y = transpose_474)[name = tensor("hidden_states_579_cast_fp16")]; + tensor var_3949_perm_0 = const()[name = tensor("op_3949_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_3953 = const()[name = tensor("op_3953"), val = tensor([1, -1, 1280])]; + tensor transpose_473 = transpose(perm = var_3949_perm_0, x = hidden_states_579_cast_fp16)[name = tensor("transpose_473")]; + tensor hidden_states_581_cast_fp16 = reshape(shape = var_3953, x = transpose_473)[name = tensor("hidden_states_581_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(684773120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(686001984))), name = tensor("mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(686002176)))]; + tensor linear_277_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_581_cast_fp16)[name = tensor("linear_277_cast_fp16")]; + tensor input_685_cast_fp16 = add(x = linear_277_cast_fp16, y = input_677_cast_fp16)[name = tensor("input_685_cast_fp16")]; + tensor input_687_axes_0 = const()[name = tensor("input_687_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_1_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(686004800)))]; + tensor mid_block_attentions_0_transformer_blocks_1_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(686007424)))]; + tensor input_687_cast_fp16 = layer_norm(axes = input_687_axes_0, beta = mid_block_attentions_0_transformer_blocks_1_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_1_norm3_weight_to_fp16, x = input_685_cast_fp16)[name = tensor("input_687_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(686010048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(695840512))), name = tensor("mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(695840704))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(695848448))), name = tensor("mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_278_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_1_ff_net_0_proj_weight_to_fp16_palettized, x = input_687_cast_fp16)[name = tensor("linear_278_cast_fp16")]; + tensor var_3975_split_sizes_0 = const()[name = tensor("op_3975_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_3975_axis_0 = const()[name = tensor("op_3975_axis_0"), val = tensor(-1)]; + tensor var_3975_cast_fp16_0, tensor var_3975_cast_fp16_1 = split(axis = var_3975_axis_0, split_sizes = var_3975_split_sizes_0, x = linear_278_cast_fp16)[name = tensor("op_3975_cast_fp16")]; + tensor var_3977_mode_0 = const()[name = tensor("op_3977_mode_0"), val = tensor("EXACT")]; + tensor var_3977_cast_fp16 = gelu(mode = var_3977_mode_0, x = var_3975_cast_fp16_1)[name = tensor("op_3977_cast_fp16")]; + tensor input_689_cast_fp16 = mul(x = var_3975_cast_fp16_0, y = var_3977_cast_fp16)[name = tensor("input_689_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(695848640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(700763904))), name = tensor("mid_block_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(700764096)))]; + tensor linear_279_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_1_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_1_ff_net_2_weight_to_fp16_palettized, x = input_689_cast_fp16)[name = tensor("linear_279_cast_fp16")]; + tensor hidden_states_589_cast_fp16 = add(x = linear_279_cast_fp16, y = input_685_cast_fp16)[name = tensor("hidden_states_589_cast_fp16")]; + tensor hidden_states_591_axes_0 = const()[name = tensor("hidden_states_591_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_2_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(700766720)))]; + tensor mid_block_attentions_0_transformer_blocks_2_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(700769344)))]; + tensor hidden_states_591_cast_fp16 = layer_norm(axes = hidden_states_591_axes_0, beta = mid_block_attentions_0_transformer_blocks_2_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_2_norm1_weight_to_fp16, x = hidden_states_589_cast_fp16)[name = tensor("hidden_states_591_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(700771968))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(702000832))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_280_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_591_cast_fp16)[name = tensor("linear_280_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(702001024))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(703229888))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_281_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_591_cast_fp16)[name = tensor("linear_281_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(703230080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(704458944))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_282_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_591_cast_fp16)[name = tensor("linear_282_cast_fp16")]; + tensor var_4012 = const()[name = tensor("op_4012"), val = tensor([1, -1, 20, 64])]; + tensor var_4013_cast_fp16 = reshape(shape = var_4012, x = linear_280_cast_fp16)[name = tensor("op_4013_cast_fp16")]; + tensor var_4015 = const()[name = tensor("op_4015"), val = tensor([1, -1, 20, 64])]; + tensor var_4016_cast_fp16 = reshape(shape = var_4015, x = linear_281_cast_fp16)[name = tensor("op_4016_cast_fp16")]; + tensor var_4018 = const()[name = tensor("op_4018"), val = tensor([1, -1, 20, 64])]; + tensor var_4019_cast_fp16 = reshape(shape = var_4018, x = linear_282_cast_fp16)[name = tensor("op_4019_cast_fp16")]; + tensor value_211_perm_0 = const()[name = tensor("value_211_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_71_y_0_to_fp16 = const()[name = tensor("mul_71_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_71_cast_fp16 = mul(x = var_4013_cast_fp16, y = mul_71_y_0_to_fp16)[name = tensor("mul_71_cast_fp16")]; + tensor matmul_52_transpose_y_0 = const()[name = tensor("matmul_52_transpose_y_0"), val = tensor(true)]; + tensor matmul_52_transpose_x_0 = const()[name = tensor("matmul_52_transpose_x_0"), val = tensor(false)]; + tensor transpose_376_perm_0 = const()[name = tensor("transpose_376_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_377_perm_0 = const()[name = tensor("transpose_377_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_471 = transpose(perm = transpose_377_perm_0, x = var_4016_cast_fp16)[name = tensor("transpose_471")]; + tensor transpose_472 = transpose(perm = transpose_376_perm_0, x = mul_71_cast_fp16)[name = tensor("transpose_472")]; + tensor matmul_52_cast_fp16 = matmul(transpose_x = matmul_52_transpose_x_0, transpose_y = matmul_52_transpose_y_0, x = transpose_472, y = transpose_471)[name = tensor("matmul_52_cast_fp16")]; + tensor softmax_52_axis_0 = const()[name = tensor("softmax_52_axis_0"), val = tensor(-1)]; + tensor softmax_52_cast_fp16 = softmax(axis = softmax_52_axis_0, x = matmul_52_cast_fp16)[name = tensor("softmax_52_cast_fp16")]; + tensor hidden_states_593_transpose_x_0 = const()[name = tensor("hidden_states_593_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_593_transpose_y_0 = const()[name = tensor("hidden_states_593_transpose_y_0"), val = tensor(false)]; + tensor transpose_470 = transpose(perm = value_211_perm_0, x = var_4019_cast_fp16)[name = tensor("transpose_470")]; + tensor hidden_states_593_cast_fp16 = matmul(transpose_x = hidden_states_593_transpose_x_0, transpose_y = hidden_states_593_transpose_y_0, x = softmax_52_cast_fp16, y = transpose_470)[name = tensor("hidden_states_593_cast_fp16")]; + tensor var_4022_perm_0 = const()[name = tensor("op_4022_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4026 = const()[name = tensor("op_4026"), val = tensor([1, -1, 1280])]; + tensor transpose_469 = transpose(perm = var_4022_perm_0, x = hidden_states_593_cast_fp16)[name = tensor("transpose_469")]; + tensor hidden_states_595_cast_fp16 = reshape(shape = var_4026, x = transpose_469)[name = tensor("hidden_states_595_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(704459136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(705688000))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(705688192)))]; + tensor linear_283_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_595_cast_fp16)[name = tensor("linear_283_cast_fp16")]; + tensor input_697_cast_fp16 = add(x = linear_283_cast_fp16, y = hidden_states_589_cast_fp16)[name = tensor("input_697_cast_fp16")]; + tensor input_699_axes_0 = const()[name = tensor("input_699_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_2_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(705690816)))]; + tensor mid_block_attentions_0_transformer_blocks_2_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(705693440)))]; + tensor input_699_cast_fp16 = layer_norm(axes = input_699_axes_0, beta = mid_block_attentions_0_transformer_blocks_2_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_2_norm2_weight_to_fp16, x = input_697_cast_fp16)[name = tensor("input_699_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(705696064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(706924928))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_284_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn2_to_q_weight_to_fp16_palettized, x = input_699_cast_fp16)[name = tensor("linear_284_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(706925120))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(708891264))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_285_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_285_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(708891456))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(710857600))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_286_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_286_cast_fp16")]; + tensor var_4058 = const()[name = tensor("op_4058"), val = tensor([1, -1, 20, 64])]; + tensor var_4059_cast_fp16 = reshape(shape = var_4058, x = linear_284_cast_fp16)[name = tensor("op_4059_cast_fp16")]; + tensor var_4061 = const()[name = tensor("op_4061"), val = tensor([1, -1, 20, 64])]; + tensor var_4062_cast_fp16 = reshape(shape = var_4061, x = linear_285_cast_fp16)[name = tensor("op_4062_cast_fp16")]; + tensor var_4064 = const()[name = tensor("op_4064"), val = tensor([1, -1, 20, 64])]; + tensor var_4065_cast_fp16 = reshape(shape = var_4064, x = linear_286_cast_fp16)[name = tensor("op_4065_cast_fp16")]; + tensor value_215_perm_0 = const()[name = tensor("value_215_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_72_y_0_to_fp16 = const()[name = tensor("mul_72_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_72_cast_fp16 = mul(x = var_4059_cast_fp16, y = mul_72_y_0_to_fp16)[name = tensor("mul_72_cast_fp16")]; + tensor matmul_53_transpose_y_0 = const()[name = tensor("matmul_53_transpose_y_0"), val = tensor(true)]; + tensor matmul_53_transpose_x_0 = const()[name = tensor("matmul_53_transpose_x_0"), val = tensor(false)]; + tensor transpose_378_perm_0 = const()[name = tensor("transpose_378_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_379_perm_0 = const()[name = tensor("transpose_379_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_467 = transpose(perm = transpose_379_perm_0, x = var_4062_cast_fp16)[name = tensor("transpose_467")]; + tensor transpose_468 = transpose(perm = transpose_378_perm_0, x = mul_72_cast_fp16)[name = tensor("transpose_468")]; + tensor matmul_53_cast_fp16 = matmul(transpose_x = matmul_53_transpose_x_0, transpose_y = matmul_53_transpose_y_0, x = transpose_468, y = transpose_467)[name = tensor("matmul_53_cast_fp16")]; + tensor softmax_53_axis_0 = const()[name = tensor("softmax_53_axis_0"), val = tensor(-1)]; + tensor softmax_53_cast_fp16 = softmax(axis = softmax_53_axis_0, x = matmul_53_cast_fp16)[name = tensor("softmax_53_cast_fp16")]; + tensor hidden_states_599_transpose_x_0 = const()[name = tensor("hidden_states_599_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_599_transpose_y_0 = const()[name = tensor("hidden_states_599_transpose_y_0"), val = tensor(false)]; + tensor transpose_466 = transpose(perm = value_215_perm_0, x = var_4065_cast_fp16)[name = tensor("transpose_466")]; + tensor hidden_states_599_cast_fp16 = matmul(transpose_x = hidden_states_599_transpose_x_0, transpose_y = hidden_states_599_transpose_y_0, x = softmax_53_cast_fp16, y = transpose_466)[name = tensor("hidden_states_599_cast_fp16")]; + tensor var_4068_perm_0 = const()[name = tensor("op_4068_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4072 = const()[name = tensor("op_4072"), val = tensor([1, -1, 1280])]; + tensor transpose_465 = transpose(perm = var_4068_perm_0, x = hidden_states_599_cast_fp16)[name = tensor("transpose_465")]; + tensor hidden_states_601_cast_fp16 = reshape(shape = var_4072, x = transpose_465)[name = tensor("hidden_states_601_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(710857792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(712086656))), name = tensor("mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(712086848)))]; + tensor linear_287_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_601_cast_fp16)[name = tensor("linear_287_cast_fp16")]; + tensor input_705_cast_fp16 = add(x = linear_287_cast_fp16, y = input_697_cast_fp16)[name = tensor("input_705_cast_fp16")]; + tensor input_707_axes_0 = const()[name = tensor("input_707_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_2_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(712089472)))]; + tensor mid_block_attentions_0_transformer_blocks_2_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(712092096)))]; + tensor input_707_cast_fp16 = layer_norm(axes = input_707_axes_0, beta = mid_block_attentions_0_transformer_blocks_2_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_2_norm3_weight_to_fp16, x = input_705_cast_fp16)[name = tensor("input_707_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(712094720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(721925184))), name = tensor("mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(721925376))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(721933120))), name = tensor("mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_288_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_2_ff_net_0_proj_weight_to_fp16_palettized, x = input_707_cast_fp16)[name = tensor("linear_288_cast_fp16")]; + tensor var_4094_split_sizes_0 = const()[name = tensor("op_4094_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4094_axis_0 = const()[name = tensor("op_4094_axis_0"), val = tensor(-1)]; + tensor var_4094_cast_fp16_0, tensor var_4094_cast_fp16_1 = split(axis = var_4094_axis_0, split_sizes = var_4094_split_sizes_0, x = linear_288_cast_fp16)[name = tensor("op_4094_cast_fp16")]; + tensor var_4096_mode_0 = const()[name = tensor("op_4096_mode_0"), val = tensor("EXACT")]; + tensor var_4096_cast_fp16 = gelu(mode = var_4096_mode_0, x = var_4094_cast_fp16_1)[name = tensor("op_4096_cast_fp16")]; + tensor input_709_cast_fp16 = mul(x = var_4094_cast_fp16_0, y = var_4096_cast_fp16)[name = tensor("input_709_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(721933312))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(726848576))), name = tensor("mid_block_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(726848768)))]; + tensor linear_289_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_2_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_2_ff_net_2_weight_to_fp16_palettized, x = input_709_cast_fp16)[name = tensor("linear_289_cast_fp16")]; + tensor hidden_states_609_cast_fp16 = add(x = linear_289_cast_fp16, y = input_705_cast_fp16)[name = tensor("hidden_states_609_cast_fp16")]; + tensor hidden_states_611_axes_0 = const()[name = tensor("hidden_states_611_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_3_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(726851392)))]; + tensor mid_block_attentions_0_transformer_blocks_3_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(726854016)))]; + tensor hidden_states_611_cast_fp16 = layer_norm(axes = hidden_states_611_axes_0, beta = mid_block_attentions_0_transformer_blocks_3_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_3_norm1_weight_to_fp16, x = hidden_states_609_cast_fp16)[name = tensor("hidden_states_611_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(726856640))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(728085504))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_290_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_611_cast_fp16)[name = tensor("linear_290_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(728085696))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(729314560))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_291_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_611_cast_fp16)[name = tensor("linear_291_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(729314752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(730543616))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_292_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_611_cast_fp16)[name = tensor("linear_292_cast_fp16")]; + tensor var_4131 = const()[name = tensor("op_4131"), val = tensor([1, -1, 20, 64])]; + tensor var_4132_cast_fp16 = reshape(shape = var_4131, x = linear_290_cast_fp16)[name = tensor("op_4132_cast_fp16")]; + tensor var_4134 = const()[name = tensor("op_4134"), val = tensor([1, -1, 20, 64])]; + tensor var_4135_cast_fp16 = reshape(shape = var_4134, x = linear_291_cast_fp16)[name = tensor("op_4135_cast_fp16")]; + tensor var_4137 = const()[name = tensor("op_4137"), val = tensor([1, -1, 20, 64])]; + tensor var_4138_cast_fp16 = reshape(shape = var_4137, x = linear_292_cast_fp16)[name = tensor("op_4138_cast_fp16")]; + tensor value_219_perm_0 = const()[name = tensor("value_219_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_73_y_0_to_fp16 = const()[name = tensor("mul_73_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_73_cast_fp16 = mul(x = var_4132_cast_fp16, y = mul_73_y_0_to_fp16)[name = tensor("mul_73_cast_fp16")]; + tensor matmul_54_transpose_y_0 = const()[name = tensor("matmul_54_transpose_y_0"), val = tensor(true)]; + tensor matmul_54_transpose_x_0 = const()[name = tensor("matmul_54_transpose_x_0"), val = tensor(false)]; + tensor transpose_380_perm_0 = const()[name = tensor("transpose_380_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_381_perm_0 = const()[name = tensor("transpose_381_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_463 = transpose(perm = transpose_381_perm_0, x = var_4135_cast_fp16)[name = tensor("transpose_463")]; + tensor transpose_464 = transpose(perm = transpose_380_perm_0, x = mul_73_cast_fp16)[name = tensor("transpose_464")]; + tensor matmul_54_cast_fp16 = matmul(transpose_x = matmul_54_transpose_x_0, transpose_y = matmul_54_transpose_y_0, x = transpose_464, y = transpose_463)[name = tensor("matmul_54_cast_fp16")]; + tensor softmax_54_axis_0 = const()[name = tensor("softmax_54_axis_0"), val = tensor(-1)]; + tensor softmax_54_cast_fp16 = softmax(axis = softmax_54_axis_0, x = matmul_54_cast_fp16)[name = tensor("softmax_54_cast_fp16")]; + tensor hidden_states_613_transpose_x_0 = const()[name = tensor("hidden_states_613_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_613_transpose_y_0 = const()[name = tensor("hidden_states_613_transpose_y_0"), val = tensor(false)]; + tensor transpose_462 = transpose(perm = value_219_perm_0, x = var_4138_cast_fp16)[name = tensor("transpose_462")]; + tensor hidden_states_613_cast_fp16 = matmul(transpose_x = hidden_states_613_transpose_x_0, transpose_y = hidden_states_613_transpose_y_0, x = softmax_54_cast_fp16, y = transpose_462)[name = tensor("hidden_states_613_cast_fp16")]; + tensor var_4141_perm_0 = const()[name = tensor("op_4141_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4145 = const()[name = tensor("op_4145"), val = tensor([1, -1, 1280])]; + tensor transpose_461 = transpose(perm = var_4141_perm_0, x = hidden_states_613_cast_fp16)[name = tensor("transpose_461")]; + tensor hidden_states_615_cast_fp16 = reshape(shape = var_4145, x = transpose_461)[name = tensor("hidden_states_615_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(730543808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(731772672))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(731772864)))]; + tensor linear_293_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_615_cast_fp16)[name = tensor("linear_293_cast_fp16")]; + tensor input_717_cast_fp16 = add(x = linear_293_cast_fp16, y = hidden_states_609_cast_fp16)[name = tensor("input_717_cast_fp16")]; + tensor input_719_axes_0 = const()[name = tensor("input_719_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_3_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(731775488)))]; + tensor mid_block_attentions_0_transformer_blocks_3_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(731778112)))]; + tensor input_719_cast_fp16 = layer_norm(axes = input_719_axes_0, beta = mid_block_attentions_0_transformer_blocks_3_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_3_norm2_weight_to_fp16, x = input_717_cast_fp16)[name = tensor("input_719_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(731780736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(733009600))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_294_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn2_to_q_weight_to_fp16_palettized, x = input_719_cast_fp16)[name = tensor("linear_294_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(733009792))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(734975936))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_295_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_295_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(734976128))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(736942272))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_296_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_296_cast_fp16")]; + tensor var_4177 = const()[name = tensor("op_4177"), val = tensor([1, -1, 20, 64])]; + tensor var_4178_cast_fp16 = reshape(shape = var_4177, x = linear_294_cast_fp16)[name = tensor("op_4178_cast_fp16")]; + tensor var_4180 = const()[name = tensor("op_4180"), val = tensor([1, -1, 20, 64])]; + tensor var_4181_cast_fp16 = reshape(shape = var_4180, x = linear_295_cast_fp16)[name = tensor("op_4181_cast_fp16")]; + tensor var_4183 = const()[name = tensor("op_4183"), val = tensor([1, -1, 20, 64])]; + tensor var_4184_cast_fp16 = reshape(shape = var_4183, x = linear_296_cast_fp16)[name = tensor("op_4184_cast_fp16")]; + tensor value_223_perm_0 = const()[name = tensor("value_223_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_74_y_0_to_fp16 = const()[name = tensor("mul_74_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_74_cast_fp16 = mul(x = var_4178_cast_fp16, y = mul_74_y_0_to_fp16)[name = tensor("mul_74_cast_fp16")]; + tensor matmul_55_transpose_y_0 = const()[name = tensor("matmul_55_transpose_y_0"), val = tensor(true)]; + tensor matmul_55_transpose_x_0 = const()[name = tensor("matmul_55_transpose_x_0"), val = tensor(false)]; + tensor transpose_382_perm_0 = const()[name = tensor("transpose_382_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_383_perm_0 = const()[name = tensor("transpose_383_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_459 = transpose(perm = transpose_383_perm_0, x = var_4181_cast_fp16)[name = tensor("transpose_459")]; + tensor transpose_460 = transpose(perm = transpose_382_perm_0, x = mul_74_cast_fp16)[name = tensor("transpose_460")]; + tensor matmul_55_cast_fp16 = matmul(transpose_x = matmul_55_transpose_x_0, transpose_y = matmul_55_transpose_y_0, x = transpose_460, y = transpose_459)[name = tensor("matmul_55_cast_fp16")]; + tensor softmax_55_axis_0 = const()[name = tensor("softmax_55_axis_0"), val = tensor(-1)]; + tensor softmax_55_cast_fp16 = softmax(axis = softmax_55_axis_0, x = matmul_55_cast_fp16)[name = tensor("softmax_55_cast_fp16")]; + tensor hidden_states_619_transpose_x_0 = const()[name = tensor("hidden_states_619_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_619_transpose_y_0 = const()[name = tensor("hidden_states_619_transpose_y_0"), val = tensor(false)]; + tensor transpose_458 = transpose(perm = value_223_perm_0, x = var_4184_cast_fp16)[name = tensor("transpose_458")]; + tensor hidden_states_619_cast_fp16 = matmul(transpose_x = hidden_states_619_transpose_x_0, transpose_y = hidden_states_619_transpose_y_0, x = softmax_55_cast_fp16, y = transpose_458)[name = tensor("hidden_states_619_cast_fp16")]; + tensor var_4187_perm_0 = const()[name = tensor("op_4187_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4191 = const()[name = tensor("op_4191"), val = tensor([1, -1, 1280])]; + tensor transpose_457 = transpose(perm = var_4187_perm_0, x = hidden_states_619_cast_fp16)[name = tensor("transpose_457")]; + tensor hidden_states_621_cast_fp16 = reshape(shape = var_4191, x = transpose_457)[name = tensor("hidden_states_621_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(736942464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(738171328))), name = tensor("mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(738171520)))]; + tensor linear_297_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_621_cast_fp16)[name = tensor("linear_297_cast_fp16")]; + tensor input_725_cast_fp16 = add(x = linear_297_cast_fp16, y = input_717_cast_fp16)[name = tensor("input_725_cast_fp16")]; + tensor input_727_axes_0 = const()[name = tensor("input_727_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_3_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(738174144)))]; + tensor mid_block_attentions_0_transformer_blocks_3_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(738176768)))]; + tensor input_727_cast_fp16 = layer_norm(axes = input_727_axes_0, beta = mid_block_attentions_0_transformer_blocks_3_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_3_norm3_weight_to_fp16, x = input_725_cast_fp16)[name = tensor("input_727_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(738179392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(748009856))), name = tensor("mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(748010048))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(748017792))), name = tensor("mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_298_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_3_ff_net_0_proj_weight_to_fp16_palettized, x = input_727_cast_fp16)[name = tensor("linear_298_cast_fp16")]; + tensor var_4213_split_sizes_0 = const()[name = tensor("op_4213_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4213_axis_0 = const()[name = tensor("op_4213_axis_0"), val = tensor(-1)]; + tensor var_4213_cast_fp16_0, tensor var_4213_cast_fp16_1 = split(axis = var_4213_axis_0, split_sizes = var_4213_split_sizes_0, x = linear_298_cast_fp16)[name = tensor("op_4213_cast_fp16")]; + tensor var_4215_mode_0 = const()[name = tensor("op_4215_mode_0"), val = tensor("EXACT")]; + tensor var_4215_cast_fp16 = gelu(mode = var_4215_mode_0, x = var_4213_cast_fp16_1)[name = tensor("op_4215_cast_fp16")]; + tensor input_729_cast_fp16 = mul(x = var_4213_cast_fp16_0, y = var_4215_cast_fp16)[name = tensor("input_729_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(748017984))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(752933248))), name = tensor("mid_block_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(752933440)))]; + tensor linear_299_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_3_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_3_ff_net_2_weight_to_fp16_palettized, x = input_729_cast_fp16)[name = tensor("linear_299_cast_fp16")]; + tensor hidden_states_629_cast_fp16 = add(x = linear_299_cast_fp16, y = input_725_cast_fp16)[name = tensor("hidden_states_629_cast_fp16")]; + tensor hidden_states_631_axes_0 = const()[name = tensor("hidden_states_631_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_4_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(752936064)))]; + tensor mid_block_attentions_0_transformer_blocks_4_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(752938688)))]; + tensor hidden_states_631_cast_fp16 = layer_norm(axes = hidden_states_631_axes_0, beta = mid_block_attentions_0_transformer_blocks_4_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_4_norm1_weight_to_fp16, x = hidden_states_629_cast_fp16)[name = tensor("hidden_states_631_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(752941312))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(754170176))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_300_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_631_cast_fp16)[name = tensor("linear_300_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(754170368))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(755399232))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_301_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_631_cast_fp16)[name = tensor("linear_301_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(755399424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(756628288))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_302_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_631_cast_fp16)[name = tensor("linear_302_cast_fp16")]; + tensor var_4250 = const()[name = tensor("op_4250"), val = tensor([1, -1, 20, 64])]; + tensor var_4251_cast_fp16 = reshape(shape = var_4250, x = linear_300_cast_fp16)[name = tensor("op_4251_cast_fp16")]; + tensor var_4253 = const()[name = tensor("op_4253"), val = tensor([1, -1, 20, 64])]; + tensor var_4254_cast_fp16 = reshape(shape = var_4253, x = linear_301_cast_fp16)[name = tensor("op_4254_cast_fp16")]; + tensor var_4256 = const()[name = tensor("op_4256"), val = tensor([1, -1, 20, 64])]; + tensor var_4257_cast_fp16 = reshape(shape = var_4256, x = linear_302_cast_fp16)[name = tensor("op_4257_cast_fp16")]; + tensor value_227_perm_0 = const()[name = tensor("value_227_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_75_y_0_to_fp16 = const()[name = tensor("mul_75_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_75_cast_fp16 = mul(x = var_4251_cast_fp16, y = mul_75_y_0_to_fp16)[name = tensor("mul_75_cast_fp16")]; + tensor matmul_56_transpose_y_0 = const()[name = tensor("matmul_56_transpose_y_0"), val = tensor(true)]; + tensor matmul_56_transpose_x_0 = const()[name = tensor("matmul_56_transpose_x_0"), val = tensor(false)]; + tensor transpose_384_perm_0 = const()[name = tensor("transpose_384_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_385_perm_0 = const()[name = tensor("transpose_385_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_455 = transpose(perm = transpose_385_perm_0, x = var_4254_cast_fp16)[name = tensor("transpose_455")]; + tensor transpose_456 = transpose(perm = transpose_384_perm_0, x = mul_75_cast_fp16)[name = tensor("transpose_456")]; + tensor matmul_56_cast_fp16 = matmul(transpose_x = matmul_56_transpose_x_0, transpose_y = matmul_56_transpose_y_0, x = transpose_456, y = transpose_455)[name = tensor("matmul_56_cast_fp16")]; + tensor softmax_56_axis_0 = const()[name = tensor("softmax_56_axis_0"), val = tensor(-1)]; + tensor softmax_56_cast_fp16 = softmax(axis = softmax_56_axis_0, x = matmul_56_cast_fp16)[name = tensor("softmax_56_cast_fp16")]; + tensor hidden_states_633_transpose_x_0 = const()[name = tensor("hidden_states_633_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_633_transpose_y_0 = const()[name = tensor("hidden_states_633_transpose_y_0"), val = tensor(false)]; + tensor transpose_454 = transpose(perm = value_227_perm_0, x = var_4257_cast_fp16)[name = tensor("transpose_454")]; + tensor hidden_states_633_cast_fp16 = matmul(transpose_x = hidden_states_633_transpose_x_0, transpose_y = hidden_states_633_transpose_y_0, x = softmax_56_cast_fp16, y = transpose_454)[name = tensor("hidden_states_633_cast_fp16")]; + tensor var_4260_perm_0 = const()[name = tensor("op_4260_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4264 = const()[name = tensor("op_4264"), val = tensor([1, -1, 1280])]; + tensor transpose_453 = transpose(perm = var_4260_perm_0, x = hidden_states_633_cast_fp16)[name = tensor("transpose_453")]; + tensor hidden_states_635_cast_fp16 = reshape(shape = var_4264, x = transpose_453)[name = tensor("hidden_states_635_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(756628480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(757857344))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(757857536)))]; + tensor linear_303_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_635_cast_fp16)[name = tensor("linear_303_cast_fp16")]; + tensor input_737_cast_fp16 = add(x = linear_303_cast_fp16, y = hidden_states_629_cast_fp16)[name = tensor("input_737_cast_fp16")]; + tensor input_739_axes_0 = const()[name = tensor("input_739_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_4_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(757860160)))]; + tensor mid_block_attentions_0_transformer_blocks_4_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(757862784)))]; + tensor input_739_cast_fp16 = layer_norm(axes = input_739_axes_0, beta = mid_block_attentions_0_transformer_blocks_4_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_4_norm2_weight_to_fp16, x = input_737_cast_fp16)[name = tensor("input_739_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(757865408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(759094272))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_304_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn2_to_q_weight_to_fp16_palettized, x = input_739_cast_fp16)[name = tensor("linear_304_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(759094464))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(761060608))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_305_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_305_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(761060800))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(763026944))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_306_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_306_cast_fp16")]; + tensor var_4296 = const()[name = tensor("op_4296"), val = tensor([1, -1, 20, 64])]; + tensor var_4297_cast_fp16 = reshape(shape = var_4296, x = linear_304_cast_fp16)[name = tensor("op_4297_cast_fp16")]; + tensor var_4299 = const()[name = tensor("op_4299"), val = tensor([1, -1, 20, 64])]; + tensor var_4300_cast_fp16 = reshape(shape = var_4299, x = linear_305_cast_fp16)[name = tensor("op_4300_cast_fp16")]; + tensor var_4302 = const()[name = tensor("op_4302"), val = tensor([1, -1, 20, 64])]; + tensor var_4303_cast_fp16 = reshape(shape = var_4302, x = linear_306_cast_fp16)[name = tensor("op_4303_cast_fp16")]; + tensor value_231_perm_0 = const()[name = tensor("value_231_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_76_y_0_to_fp16 = const()[name = tensor("mul_76_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_76_cast_fp16 = mul(x = var_4297_cast_fp16, y = mul_76_y_0_to_fp16)[name = tensor("mul_76_cast_fp16")]; + tensor matmul_57_transpose_y_0 = const()[name = tensor("matmul_57_transpose_y_0"), val = tensor(true)]; + tensor matmul_57_transpose_x_0 = const()[name = tensor("matmul_57_transpose_x_0"), val = tensor(false)]; + tensor transpose_386_perm_0 = const()[name = tensor("transpose_386_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_387_perm_0 = const()[name = tensor("transpose_387_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_451 = transpose(perm = transpose_387_perm_0, x = var_4300_cast_fp16)[name = tensor("transpose_451")]; + tensor transpose_452 = transpose(perm = transpose_386_perm_0, x = mul_76_cast_fp16)[name = tensor("transpose_452")]; + tensor matmul_57_cast_fp16 = matmul(transpose_x = matmul_57_transpose_x_0, transpose_y = matmul_57_transpose_y_0, x = transpose_452, y = transpose_451)[name = tensor("matmul_57_cast_fp16")]; + tensor softmax_57_axis_0 = const()[name = tensor("softmax_57_axis_0"), val = tensor(-1)]; + tensor softmax_57_cast_fp16 = softmax(axis = softmax_57_axis_0, x = matmul_57_cast_fp16)[name = tensor("softmax_57_cast_fp16")]; + tensor hidden_states_639_transpose_x_0 = const()[name = tensor("hidden_states_639_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_639_transpose_y_0 = const()[name = tensor("hidden_states_639_transpose_y_0"), val = tensor(false)]; + tensor transpose_450 = transpose(perm = value_231_perm_0, x = var_4303_cast_fp16)[name = tensor("transpose_450")]; + tensor hidden_states_639_cast_fp16 = matmul(transpose_x = hidden_states_639_transpose_x_0, transpose_y = hidden_states_639_transpose_y_0, x = softmax_57_cast_fp16, y = transpose_450)[name = tensor("hidden_states_639_cast_fp16")]; + tensor var_4306_perm_0 = const()[name = tensor("op_4306_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4310 = const()[name = tensor("op_4310"), val = tensor([1, -1, 1280])]; + tensor transpose_449 = transpose(perm = var_4306_perm_0, x = hidden_states_639_cast_fp16)[name = tensor("transpose_449")]; + tensor hidden_states_641_cast_fp16 = reshape(shape = var_4310, x = transpose_449)[name = tensor("hidden_states_641_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(763027136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(764256000))), name = tensor("mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(764256192)))]; + tensor linear_307_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_641_cast_fp16)[name = tensor("linear_307_cast_fp16")]; + tensor input_745_cast_fp16 = add(x = linear_307_cast_fp16, y = input_737_cast_fp16)[name = tensor("input_745_cast_fp16")]; + tensor input_747_axes_0 = const()[name = tensor("input_747_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_4_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(764258816)))]; + tensor mid_block_attentions_0_transformer_blocks_4_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(764261440)))]; + tensor input_747_cast_fp16 = layer_norm(axes = input_747_axes_0, beta = mid_block_attentions_0_transformer_blocks_4_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_4_norm3_weight_to_fp16, x = input_745_cast_fp16)[name = tensor("input_747_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(764264064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(774094528))), name = tensor("mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(774094720))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(774102464))), name = tensor("mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_308_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_4_ff_net_0_proj_weight_to_fp16_palettized, x = input_747_cast_fp16)[name = tensor("linear_308_cast_fp16")]; + tensor var_4332_split_sizes_0 = const()[name = tensor("op_4332_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4332_axis_0 = const()[name = tensor("op_4332_axis_0"), val = tensor(-1)]; + tensor var_4332_cast_fp16_0, tensor var_4332_cast_fp16_1 = split(axis = var_4332_axis_0, split_sizes = var_4332_split_sizes_0, x = linear_308_cast_fp16)[name = tensor("op_4332_cast_fp16")]; + tensor var_4334_mode_0 = const()[name = tensor("op_4334_mode_0"), val = tensor("EXACT")]; + tensor var_4334_cast_fp16 = gelu(mode = var_4334_mode_0, x = var_4332_cast_fp16_1)[name = tensor("op_4334_cast_fp16")]; + tensor input_749_cast_fp16 = mul(x = var_4332_cast_fp16_0, y = var_4334_cast_fp16)[name = tensor("input_749_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(774102656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(779017920))), name = tensor("mid_block_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(779018112)))]; + tensor linear_309_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_4_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_4_ff_net_2_weight_to_fp16_palettized, x = input_749_cast_fp16)[name = tensor("linear_309_cast_fp16")]; + tensor hidden_states_649_cast_fp16 = add(x = linear_309_cast_fp16, y = input_745_cast_fp16)[name = tensor("hidden_states_649_cast_fp16")]; + tensor hidden_states_651_axes_0 = const()[name = tensor("hidden_states_651_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_5_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(779020736)))]; + tensor mid_block_attentions_0_transformer_blocks_5_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(779023360)))]; + tensor hidden_states_651_cast_fp16 = layer_norm(axes = hidden_states_651_axes_0, beta = mid_block_attentions_0_transformer_blocks_5_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_5_norm1_weight_to_fp16, x = hidden_states_649_cast_fp16)[name = tensor("hidden_states_651_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(779025984))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(780254848))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_310_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_651_cast_fp16)[name = tensor("linear_310_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(780255040))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(781483904))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_311_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_651_cast_fp16)[name = tensor("linear_311_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(781484096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(782712960))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_312_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_651_cast_fp16)[name = tensor("linear_312_cast_fp16")]; + tensor var_4369 = const()[name = tensor("op_4369"), val = tensor([1, -1, 20, 64])]; + tensor var_4370_cast_fp16 = reshape(shape = var_4369, x = linear_310_cast_fp16)[name = tensor("op_4370_cast_fp16")]; + tensor var_4372 = const()[name = tensor("op_4372"), val = tensor([1, -1, 20, 64])]; + tensor var_4373_cast_fp16 = reshape(shape = var_4372, x = linear_311_cast_fp16)[name = tensor("op_4373_cast_fp16")]; + tensor var_4375 = const()[name = tensor("op_4375"), val = tensor([1, -1, 20, 64])]; + tensor var_4376_cast_fp16 = reshape(shape = var_4375, x = linear_312_cast_fp16)[name = tensor("op_4376_cast_fp16")]; + tensor value_235_perm_0 = const()[name = tensor("value_235_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_77_y_0_to_fp16 = const()[name = tensor("mul_77_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_77_cast_fp16 = mul(x = var_4370_cast_fp16, y = mul_77_y_0_to_fp16)[name = tensor("mul_77_cast_fp16")]; + tensor matmul_58_transpose_y_0 = const()[name = tensor("matmul_58_transpose_y_0"), val = tensor(true)]; + tensor matmul_58_transpose_x_0 = const()[name = tensor("matmul_58_transpose_x_0"), val = tensor(false)]; + tensor transpose_388_perm_0 = const()[name = tensor("transpose_388_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_389_perm_0 = const()[name = tensor("transpose_389_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_447 = transpose(perm = transpose_389_perm_0, x = var_4373_cast_fp16)[name = tensor("transpose_447")]; + tensor transpose_448 = transpose(perm = transpose_388_perm_0, x = mul_77_cast_fp16)[name = tensor("transpose_448")]; + tensor matmul_58_cast_fp16 = matmul(transpose_x = matmul_58_transpose_x_0, transpose_y = matmul_58_transpose_y_0, x = transpose_448, y = transpose_447)[name = tensor("matmul_58_cast_fp16")]; + tensor softmax_58_axis_0 = const()[name = tensor("softmax_58_axis_0"), val = tensor(-1)]; + tensor softmax_58_cast_fp16 = softmax(axis = softmax_58_axis_0, x = matmul_58_cast_fp16)[name = tensor("softmax_58_cast_fp16")]; + tensor hidden_states_653_transpose_x_0 = const()[name = tensor("hidden_states_653_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_653_transpose_y_0 = const()[name = tensor("hidden_states_653_transpose_y_0"), val = tensor(false)]; + tensor transpose_446 = transpose(perm = value_235_perm_0, x = var_4376_cast_fp16)[name = tensor("transpose_446")]; + tensor hidden_states_653_cast_fp16 = matmul(transpose_x = hidden_states_653_transpose_x_0, transpose_y = hidden_states_653_transpose_y_0, x = softmax_58_cast_fp16, y = transpose_446)[name = tensor("hidden_states_653_cast_fp16")]; + tensor var_4379_perm_0 = const()[name = tensor("op_4379_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4383 = const()[name = tensor("op_4383"), val = tensor([1, -1, 1280])]; + tensor transpose_445 = transpose(perm = var_4379_perm_0, x = hidden_states_653_cast_fp16)[name = tensor("transpose_445")]; + tensor hidden_states_655_cast_fp16 = reshape(shape = var_4383, x = transpose_445)[name = tensor("hidden_states_655_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(782713152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783942016))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783942208)))]; + tensor linear_313_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_655_cast_fp16)[name = tensor("linear_313_cast_fp16")]; + tensor input_757_cast_fp16 = add(x = linear_313_cast_fp16, y = hidden_states_649_cast_fp16)[name = tensor("input_757_cast_fp16")]; + tensor input_759_axes_0 = const()[name = tensor("input_759_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_5_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783944832)))]; + tensor mid_block_attentions_0_transformer_blocks_5_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783947456)))]; + tensor input_759_cast_fp16 = layer_norm(axes = input_759_axes_0, beta = mid_block_attentions_0_transformer_blocks_5_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_5_norm2_weight_to_fp16, x = input_757_cast_fp16)[name = tensor("input_759_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(783950080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(785178944))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_314_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn2_to_q_weight_to_fp16_palettized, x = input_759_cast_fp16)[name = tensor("linear_314_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(785179136))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(787145280))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_315_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_315_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(787145472))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(789111616))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_316_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_316_cast_fp16")]; + tensor var_4415 = const()[name = tensor("op_4415"), val = tensor([1, -1, 20, 64])]; + tensor var_4416_cast_fp16 = reshape(shape = var_4415, x = linear_314_cast_fp16)[name = tensor("op_4416_cast_fp16")]; + tensor var_4418 = const()[name = tensor("op_4418"), val = tensor([1, -1, 20, 64])]; + tensor var_4419_cast_fp16 = reshape(shape = var_4418, x = linear_315_cast_fp16)[name = tensor("op_4419_cast_fp16")]; + tensor var_4421 = const()[name = tensor("op_4421"), val = tensor([1, -1, 20, 64])]; + tensor var_4422_cast_fp16 = reshape(shape = var_4421, x = linear_316_cast_fp16)[name = tensor("op_4422_cast_fp16")]; + tensor value_239_perm_0 = const()[name = tensor("value_239_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_78_y_0_to_fp16 = const()[name = tensor("mul_78_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_78_cast_fp16 = mul(x = var_4416_cast_fp16, y = mul_78_y_0_to_fp16)[name = tensor("mul_78_cast_fp16")]; + tensor matmul_59_transpose_y_0 = const()[name = tensor("matmul_59_transpose_y_0"), val = tensor(true)]; + tensor matmul_59_transpose_x_0 = const()[name = tensor("matmul_59_transpose_x_0"), val = tensor(false)]; + tensor transpose_390_perm_0 = const()[name = tensor("transpose_390_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_391_perm_0 = const()[name = tensor("transpose_391_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_443 = transpose(perm = transpose_391_perm_0, x = var_4419_cast_fp16)[name = tensor("transpose_443")]; + tensor transpose_444 = transpose(perm = transpose_390_perm_0, x = mul_78_cast_fp16)[name = tensor("transpose_444")]; + tensor matmul_59_cast_fp16 = matmul(transpose_x = matmul_59_transpose_x_0, transpose_y = matmul_59_transpose_y_0, x = transpose_444, y = transpose_443)[name = tensor("matmul_59_cast_fp16")]; + tensor softmax_59_axis_0 = const()[name = tensor("softmax_59_axis_0"), val = tensor(-1)]; + tensor softmax_59_cast_fp16 = softmax(axis = softmax_59_axis_0, x = matmul_59_cast_fp16)[name = tensor("softmax_59_cast_fp16")]; + tensor hidden_states_659_transpose_x_0 = const()[name = tensor("hidden_states_659_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_659_transpose_y_0 = const()[name = tensor("hidden_states_659_transpose_y_0"), val = tensor(false)]; + tensor transpose_442 = transpose(perm = value_239_perm_0, x = var_4422_cast_fp16)[name = tensor("transpose_442")]; + tensor hidden_states_659_cast_fp16 = matmul(transpose_x = hidden_states_659_transpose_x_0, transpose_y = hidden_states_659_transpose_y_0, x = softmax_59_cast_fp16, y = transpose_442)[name = tensor("hidden_states_659_cast_fp16")]; + tensor var_4425_perm_0 = const()[name = tensor("op_4425_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4429 = const()[name = tensor("op_4429"), val = tensor([1, -1, 1280])]; + tensor transpose_441 = transpose(perm = var_4425_perm_0, x = hidden_states_659_cast_fp16)[name = tensor("transpose_441")]; + tensor hidden_states_661_cast_fp16 = reshape(shape = var_4429, x = transpose_441)[name = tensor("hidden_states_661_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(789111808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(790340672))), name = tensor("mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(790340864)))]; + tensor linear_317_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_661_cast_fp16)[name = tensor("linear_317_cast_fp16")]; + tensor input_765_cast_fp16 = add(x = linear_317_cast_fp16, y = input_757_cast_fp16)[name = tensor("input_765_cast_fp16")]; + tensor input_767_axes_0 = const()[name = tensor("input_767_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_5_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(790343488)))]; + tensor mid_block_attentions_0_transformer_blocks_5_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(790346112)))]; + tensor input_767_cast_fp16 = layer_norm(axes = input_767_axes_0, beta = mid_block_attentions_0_transformer_blocks_5_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_5_norm3_weight_to_fp16, x = input_765_cast_fp16)[name = tensor("input_767_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(790348736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(800179200))), name = tensor("mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(800179392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(800187136))), name = tensor("mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_318_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_5_ff_net_0_proj_weight_to_fp16_palettized, x = input_767_cast_fp16)[name = tensor("linear_318_cast_fp16")]; + tensor var_4451_split_sizes_0 = const()[name = tensor("op_4451_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4451_axis_0 = const()[name = tensor("op_4451_axis_0"), val = tensor(-1)]; + tensor var_4451_cast_fp16_0, tensor var_4451_cast_fp16_1 = split(axis = var_4451_axis_0, split_sizes = var_4451_split_sizes_0, x = linear_318_cast_fp16)[name = tensor("op_4451_cast_fp16")]; + tensor var_4453_mode_0 = const()[name = tensor("op_4453_mode_0"), val = tensor("EXACT")]; + tensor var_4453_cast_fp16 = gelu(mode = var_4453_mode_0, x = var_4451_cast_fp16_1)[name = tensor("op_4453_cast_fp16")]; + tensor input_769_cast_fp16 = mul(x = var_4451_cast_fp16_0, y = var_4453_cast_fp16)[name = tensor("input_769_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(800187328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(805102592))), name = tensor("mid_block_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(805102784)))]; + tensor linear_319_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_5_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_5_ff_net_2_weight_to_fp16_palettized, x = input_769_cast_fp16)[name = tensor("linear_319_cast_fp16")]; + tensor hidden_states_669_cast_fp16 = add(x = linear_319_cast_fp16, y = input_765_cast_fp16)[name = tensor("hidden_states_669_cast_fp16")]; + tensor hidden_states_671_axes_0 = const()[name = tensor("hidden_states_671_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_6_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(805105408)))]; + tensor mid_block_attentions_0_transformer_blocks_6_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(805108032)))]; + tensor hidden_states_671_cast_fp16 = layer_norm(axes = hidden_states_671_axes_0, beta = mid_block_attentions_0_transformer_blocks_6_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_6_norm1_weight_to_fp16, x = hidden_states_669_cast_fp16)[name = tensor("hidden_states_671_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(805110656))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(806339520))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_320_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_671_cast_fp16)[name = tensor("linear_320_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(806339712))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(807568576))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_321_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_671_cast_fp16)[name = tensor("linear_321_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(807568768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(808797632))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_322_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_671_cast_fp16)[name = tensor("linear_322_cast_fp16")]; + tensor var_4488 = const()[name = tensor("op_4488"), val = tensor([1, -1, 20, 64])]; + tensor var_4489_cast_fp16 = reshape(shape = var_4488, x = linear_320_cast_fp16)[name = tensor("op_4489_cast_fp16")]; + tensor var_4491 = const()[name = tensor("op_4491"), val = tensor([1, -1, 20, 64])]; + tensor var_4492_cast_fp16 = reshape(shape = var_4491, x = linear_321_cast_fp16)[name = tensor("op_4492_cast_fp16")]; + tensor var_4494 = const()[name = tensor("op_4494"), val = tensor([1, -1, 20, 64])]; + tensor var_4495_cast_fp16 = reshape(shape = var_4494, x = linear_322_cast_fp16)[name = tensor("op_4495_cast_fp16")]; + tensor value_243_perm_0 = const()[name = tensor("value_243_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_79_y_0_to_fp16 = const()[name = tensor("mul_79_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_79_cast_fp16 = mul(x = var_4489_cast_fp16, y = mul_79_y_0_to_fp16)[name = tensor("mul_79_cast_fp16")]; + tensor matmul_60_transpose_y_0 = const()[name = tensor("matmul_60_transpose_y_0"), val = tensor(true)]; + tensor matmul_60_transpose_x_0 = const()[name = tensor("matmul_60_transpose_x_0"), val = tensor(false)]; + tensor transpose_392_perm_0 = const()[name = tensor("transpose_392_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_393_perm_0 = const()[name = tensor("transpose_393_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_439 = transpose(perm = transpose_393_perm_0, x = var_4492_cast_fp16)[name = tensor("transpose_439")]; + tensor transpose_440 = transpose(perm = transpose_392_perm_0, x = mul_79_cast_fp16)[name = tensor("transpose_440")]; + tensor matmul_60_cast_fp16 = matmul(transpose_x = matmul_60_transpose_x_0, transpose_y = matmul_60_transpose_y_0, x = transpose_440, y = transpose_439)[name = tensor("matmul_60_cast_fp16")]; + tensor softmax_60_axis_0 = const()[name = tensor("softmax_60_axis_0"), val = tensor(-1)]; + tensor softmax_60_cast_fp16 = softmax(axis = softmax_60_axis_0, x = matmul_60_cast_fp16)[name = tensor("softmax_60_cast_fp16")]; + tensor hidden_states_673_transpose_x_0 = const()[name = tensor("hidden_states_673_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_673_transpose_y_0 = const()[name = tensor("hidden_states_673_transpose_y_0"), val = tensor(false)]; + tensor transpose_438 = transpose(perm = value_243_perm_0, x = var_4495_cast_fp16)[name = tensor("transpose_438")]; + tensor hidden_states_673_cast_fp16 = matmul(transpose_x = hidden_states_673_transpose_x_0, transpose_y = hidden_states_673_transpose_y_0, x = softmax_60_cast_fp16, y = transpose_438)[name = tensor("hidden_states_673_cast_fp16")]; + tensor var_4498_perm_0 = const()[name = tensor("op_4498_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4502 = const()[name = tensor("op_4502"), val = tensor([1, -1, 1280])]; + tensor transpose_437 = transpose(perm = var_4498_perm_0, x = hidden_states_673_cast_fp16)[name = tensor("transpose_437")]; + tensor hidden_states_675_cast_fp16 = reshape(shape = var_4502, x = transpose_437)[name = tensor("hidden_states_675_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(808797824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(810026688))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(810026880)))]; + tensor linear_323_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_675_cast_fp16)[name = tensor("linear_323_cast_fp16")]; + tensor input_777_cast_fp16 = add(x = linear_323_cast_fp16, y = hidden_states_669_cast_fp16)[name = tensor("input_777_cast_fp16")]; + tensor input_779_axes_0 = const()[name = tensor("input_779_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_6_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(810029504)))]; + tensor mid_block_attentions_0_transformer_blocks_6_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(810032128)))]; + tensor input_779_cast_fp16 = layer_norm(axes = input_779_axes_0, beta = mid_block_attentions_0_transformer_blocks_6_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_6_norm2_weight_to_fp16, x = input_777_cast_fp16)[name = tensor("input_779_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(810034752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(811263616))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_324_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn2_to_q_weight_to_fp16_palettized, x = input_779_cast_fp16)[name = tensor("linear_324_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(811263808))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(813229952))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_325_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_325_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(813230144))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(815196288))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_326_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_326_cast_fp16")]; + tensor var_4534 = const()[name = tensor("op_4534"), val = tensor([1, -1, 20, 64])]; + tensor var_4535_cast_fp16 = reshape(shape = var_4534, x = linear_324_cast_fp16)[name = tensor("op_4535_cast_fp16")]; + tensor var_4537 = const()[name = tensor("op_4537"), val = tensor([1, -1, 20, 64])]; + tensor var_4538_cast_fp16 = reshape(shape = var_4537, x = linear_325_cast_fp16)[name = tensor("op_4538_cast_fp16")]; + tensor var_4540 = const()[name = tensor("op_4540"), val = tensor([1, -1, 20, 64])]; + tensor var_4541_cast_fp16 = reshape(shape = var_4540, x = linear_326_cast_fp16)[name = tensor("op_4541_cast_fp16")]; + tensor value_247_perm_0 = const()[name = tensor("value_247_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_80_y_0_to_fp16 = const()[name = tensor("mul_80_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_80_cast_fp16 = mul(x = var_4535_cast_fp16, y = mul_80_y_0_to_fp16)[name = tensor("mul_80_cast_fp16")]; + tensor matmul_61_transpose_y_0 = const()[name = tensor("matmul_61_transpose_y_0"), val = tensor(true)]; + tensor matmul_61_transpose_x_0 = const()[name = tensor("matmul_61_transpose_x_0"), val = tensor(false)]; + tensor transpose_394_perm_0 = const()[name = tensor("transpose_394_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_395_perm_0 = const()[name = tensor("transpose_395_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_435 = transpose(perm = transpose_395_perm_0, x = var_4538_cast_fp16)[name = tensor("transpose_435")]; + tensor transpose_436 = transpose(perm = transpose_394_perm_0, x = mul_80_cast_fp16)[name = tensor("transpose_436")]; + tensor matmul_61_cast_fp16 = matmul(transpose_x = matmul_61_transpose_x_0, transpose_y = matmul_61_transpose_y_0, x = transpose_436, y = transpose_435)[name = tensor("matmul_61_cast_fp16")]; + tensor softmax_61_axis_0 = const()[name = tensor("softmax_61_axis_0"), val = tensor(-1)]; + tensor softmax_61_cast_fp16 = softmax(axis = softmax_61_axis_0, x = matmul_61_cast_fp16)[name = tensor("softmax_61_cast_fp16")]; + tensor hidden_states_679_transpose_x_0 = const()[name = tensor("hidden_states_679_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_679_transpose_y_0 = const()[name = tensor("hidden_states_679_transpose_y_0"), val = tensor(false)]; + tensor transpose_434 = transpose(perm = value_247_perm_0, x = var_4541_cast_fp16)[name = tensor("transpose_434")]; + tensor hidden_states_679_cast_fp16 = matmul(transpose_x = hidden_states_679_transpose_x_0, transpose_y = hidden_states_679_transpose_y_0, x = softmax_61_cast_fp16, y = transpose_434)[name = tensor("hidden_states_679_cast_fp16")]; + tensor var_4544_perm_0 = const()[name = tensor("op_4544_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4548 = const()[name = tensor("op_4548"), val = tensor([1, -1, 1280])]; + tensor transpose_433 = transpose(perm = var_4544_perm_0, x = hidden_states_679_cast_fp16)[name = tensor("transpose_433")]; + tensor hidden_states_681_cast_fp16 = reshape(shape = var_4548, x = transpose_433)[name = tensor("hidden_states_681_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(815196480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(816425344))), name = tensor("mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(816425536)))]; + tensor linear_327_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_681_cast_fp16)[name = tensor("linear_327_cast_fp16")]; + tensor input_785_cast_fp16 = add(x = linear_327_cast_fp16, y = input_777_cast_fp16)[name = tensor("input_785_cast_fp16")]; + tensor input_787_axes_0 = const()[name = tensor("input_787_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_6_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(816428160)))]; + tensor mid_block_attentions_0_transformer_blocks_6_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(816430784)))]; + tensor input_787_cast_fp16 = layer_norm(axes = input_787_axes_0, beta = mid_block_attentions_0_transformer_blocks_6_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_6_norm3_weight_to_fp16, x = input_785_cast_fp16)[name = tensor("input_787_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(816433408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(826263872))), name = tensor("mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(826264064))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(826271808))), name = tensor("mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_328_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_6_ff_net_0_proj_weight_to_fp16_palettized, x = input_787_cast_fp16)[name = tensor("linear_328_cast_fp16")]; + tensor var_4570_split_sizes_0 = const()[name = tensor("op_4570_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4570_axis_0 = const()[name = tensor("op_4570_axis_0"), val = tensor(-1)]; + tensor var_4570_cast_fp16_0, tensor var_4570_cast_fp16_1 = split(axis = var_4570_axis_0, split_sizes = var_4570_split_sizes_0, x = linear_328_cast_fp16)[name = tensor("op_4570_cast_fp16")]; + tensor var_4572_mode_0 = const()[name = tensor("op_4572_mode_0"), val = tensor("EXACT")]; + tensor var_4572_cast_fp16 = gelu(mode = var_4572_mode_0, x = var_4570_cast_fp16_1)[name = tensor("op_4572_cast_fp16")]; + tensor input_789_cast_fp16 = mul(x = var_4570_cast_fp16_0, y = var_4572_cast_fp16)[name = tensor("input_789_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(826272000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(831187264))), name = tensor("mid_block_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(831187456)))]; + tensor linear_329_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_6_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_6_ff_net_2_weight_to_fp16_palettized, x = input_789_cast_fp16)[name = tensor("linear_329_cast_fp16")]; + tensor hidden_states_689_cast_fp16 = add(x = linear_329_cast_fp16, y = input_785_cast_fp16)[name = tensor("hidden_states_689_cast_fp16")]; + tensor hidden_states_691_axes_0 = const()[name = tensor("hidden_states_691_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_7_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(831190080)))]; + tensor mid_block_attentions_0_transformer_blocks_7_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(831192704)))]; + tensor hidden_states_691_cast_fp16 = layer_norm(axes = hidden_states_691_axes_0, beta = mid_block_attentions_0_transformer_blocks_7_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_7_norm1_weight_to_fp16, x = hidden_states_689_cast_fp16)[name = tensor("hidden_states_691_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(831195328))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(832424192))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_330_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_691_cast_fp16)[name = tensor("linear_330_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(832424384))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(833653248))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_331_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_691_cast_fp16)[name = tensor("linear_331_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(833653440))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(834882304))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_332_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_691_cast_fp16)[name = tensor("linear_332_cast_fp16")]; + tensor var_4607 = const()[name = tensor("op_4607"), val = tensor([1, -1, 20, 64])]; + tensor var_4608_cast_fp16 = reshape(shape = var_4607, x = linear_330_cast_fp16)[name = tensor("op_4608_cast_fp16")]; + tensor var_4610 = const()[name = tensor("op_4610"), val = tensor([1, -1, 20, 64])]; + tensor var_4611_cast_fp16 = reshape(shape = var_4610, x = linear_331_cast_fp16)[name = tensor("op_4611_cast_fp16")]; + tensor var_4613 = const()[name = tensor("op_4613"), val = tensor([1, -1, 20, 64])]; + tensor var_4614_cast_fp16 = reshape(shape = var_4613, x = linear_332_cast_fp16)[name = tensor("op_4614_cast_fp16")]; + tensor value_251_perm_0 = const()[name = tensor("value_251_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_81_y_0_to_fp16 = const()[name = tensor("mul_81_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_81_cast_fp16 = mul(x = var_4608_cast_fp16, y = mul_81_y_0_to_fp16)[name = tensor("mul_81_cast_fp16")]; + tensor matmul_62_transpose_y_0 = const()[name = tensor("matmul_62_transpose_y_0"), val = tensor(true)]; + tensor matmul_62_transpose_x_0 = const()[name = tensor("matmul_62_transpose_x_0"), val = tensor(false)]; + tensor transpose_396_perm_0 = const()[name = tensor("transpose_396_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_397_perm_0 = const()[name = tensor("transpose_397_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_431 = transpose(perm = transpose_397_perm_0, x = var_4611_cast_fp16)[name = tensor("transpose_431")]; + tensor transpose_432 = transpose(perm = transpose_396_perm_0, x = mul_81_cast_fp16)[name = tensor("transpose_432")]; + tensor matmul_62_cast_fp16 = matmul(transpose_x = matmul_62_transpose_x_0, transpose_y = matmul_62_transpose_y_0, x = transpose_432, y = transpose_431)[name = tensor("matmul_62_cast_fp16")]; + tensor softmax_62_axis_0 = const()[name = tensor("softmax_62_axis_0"), val = tensor(-1)]; + tensor softmax_62_cast_fp16 = softmax(axis = softmax_62_axis_0, x = matmul_62_cast_fp16)[name = tensor("softmax_62_cast_fp16")]; + tensor hidden_states_693_transpose_x_0 = const()[name = tensor("hidden_states_693_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_693_transpose_y_0 = const()[name = tensor("hidden_states_693_transpose_y_0"), val = tensor(false)]; + tensor transpose_430 = transpose(perm = value_251_perm_0, x = var_4614_cast_fp16)[name = tensor("transpose_430")]; + tensor hidden_states_693_cast_fp16 = matmul(transpose_x = hidden_states_693_transpose_x_0, transpose_y = hidden_states_693_transpose_y_0, x = softmax_62_cast_fp16, y = transpose_430)[name = tensor("hidden_states_693_cast_fp16")]; + tensor var_4617_perm_0 = const()[name = tensor("op_4617_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4621 = const()[name = tensor("op_4621"), val = tensor([1, -1, 1280])]; + tensor transpose_429 = transpose(perm = var_4617_perm_0, x = hidden_states_693_cast_fp16)[name = tensor("transpose_429")]; + tensor hidden_states_695_cast_fp16 = reshape(shape = var_4621, x = transpose_429)[name = tensor("hidden_states_695_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(834882496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(836111360))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(836111552)))]; + tensor linear_333_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_695_cast_fp16)[name = tensor("linear_333_cast_fp16")]; + tensor input_797_cast_fp16 = add(x = linear_333_cast_fp16, y = hidden_states_689_cast_fp16)[name = tensor("input_797_cast_fp16")]; + tensor input_799_axes_0 = const()[name = tensor("input_799_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_7_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(836114176)))]; + tensor mid_block_attentions_0_transformer_blocks_7_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(836116800)))]; + tensor input_799_cast_fp16 = layer_norm(axes = input_799_axes_0, beta = mid_block_attentions_0_transformer_blocks_7_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_7_norm2_weight_to_fp16, x = input_797_cast_fp16)[name = tensor("input_799_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(836119424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(837348288))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_334_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn2_to_q_weight_to_fp16_palettized, x = input_799_cast_fp16)[name = tensor("linear_334_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(837348480))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(839314624))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_335_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_335_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(839314816))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(841280960))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_336_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_336_cast_fp16")]; + tensor var_4653 = const()[name = tensor("op_4653"), val = tensor([1, -1, 20, 64])]; + tensor var_4654_cast_fp16 = reshape(shape = var_4653, x = linear_334_cast_fp16)[name = tensor("op_4654_cast_fp16")]; + tensor var_4656 = const()[name = tensor("op_4656"), val = tensor([1, -1, 20, 64])]; + tensor var_4657_cast_fp16 = reshape(shape = var_4656, x = linear_335_cast_fp16)[name = tensor("op_4657_cast_fp16")]; + tensor var_4659 = const()[name = tensor("op_4659"), val = tensor([1, -1, 20, 64])]; + tensor var_4660_cast_fp16 = reshape(shape = var_4659, x = linear_336_cast_fp16)[name = tensor("op_4660_cast_fp16")]; + tensor value_255_perm_0 = const()[name = tensor("value_255_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_82_y_0_to_fp16 = const()[name = tensor("mul_82_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_82_cast_fp16 = mul(x = var_4654_cast_fp16, y = mul_82_y_0_to_fp16)[name = tensor("mul_82_cast_fp16")]; + tensor matmul_63_transpose_y_0 = const()[name = tensor("matmul_63_transpose_y_0"), val = tensor(true)]; + tensor matmul_63_transpose_x_0 = const()[name = tensor("matmul_63_transpose_x_0"), val = tensor(false)]; + tensor transpose_398_perm_0 = const()[name = tensor("transpose_398_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_399_perm_0 = const()[name = tensor("transpose_399_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_427 = transpose(perm = transpose_399_perm_0, x = var_4657_cast_fp16)[name = tensor("transpose_427")]; + tensor transpose_428 = transpose(perm = transpose_398_perm_0, x = mul_82_cast_fp16)[name = tensor("transpose_428")]; + tensor matmul_63_cast_fp16 = matmul(transpose_x = matmul_63_transpose_x_0, transpose_y = matmul_63_transpose_y_0, x = transpose_428, y = transpose_427)[name = tensor("matmul_63_cast_fp16")]; + tensor softmax_63_axis_0 = const()[name = tensor("softmax_63_axis_0"), val = tensor(-1)]; + tensor softmax_63_cast_fp16 = softmax(axis = softmax_63_axis_0, x = matmul_63_cast_fp16)[name = tensor("softmax_63_cast_fp16")]; + tensor hidden_states_699_transpose_x_0 = const()[name = tensor("hidden_states_699_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_699_transpose_y_0 = const()[name = tensor("hidden_states_699_transpose_y_0"), val = tensor(false)]; + tensor transpose_426 = transpose(perm = value_255_perm_0, x = var_4660_cast_fp16)[name = tensor("transpose_426")]; + tensor hidden_states_699_cast_fp16 = matmul(transpose_x = hidden_states_699_transpose_x_0, transpose_y = hidden_states_699_transpose_y_0, x = softmax_63_cast_fp16, y = transpose_426)[name = tensor("hidden_states_699_cast_fp16")]; + tensor var_4663_perm_0 = const()[name = tensor("op_4663_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4667 = const()[name = tensor("op_4667"), val = tensor([1, -1, 1280])]; + tensor transpose_425 = transpose(perm = var_4663_perm_0, x = hidden_states_699_cast_fp16)[name = tensor("transpose_425")]; + tensor hidden_states_701_cast_fp16 = reshape(shape = var_4667, x = transpose_425)[name = tensor("hidden_states_701_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(841281152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(842510016))), name = tensor("mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(842510208)))]; + tensor linear_337_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_701_cast_fp16)[name = tensor("linear_337_cast_fp16")]; + tensor input_805_cast_fp16 = add(x = linear_337_cast_fp16, y = input_797_cast_fp16)[name = tensor("input_805_cast_fp16")]; + tensor input_807_axes_0 = const()[name = tensor("input_807_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_7_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(842512832)))]; + tensor mid_block_attentions_0_transformer_blocks_7_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(842515456)))]; + tensor input_807_cast_fp16 = layer_norm(axes = input_807_axes_0, beta = mid_block_attentions_0_transformer_blocks_7_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_7_norm3_weight_to_fp16, x = input_805_cast_fp16)[name = tensor("input_807_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(842518080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(852348544))), name = tensor("mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(852348736))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(852356480))), name = tensor("mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_338_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_7_ff_net_0_proj_weight_to_fp16_palettized, x = input_807_cast_fp16)[name = tensor("linear_338_cast_fp16")]; + tensor var_4689_split_sizes_0 = const()[name = tensor("op_4689_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4689_axis_0 = const()[name = tensor("op_4689_axis_0"), val = tensor(-1)]; + tensor var_4689_cast_fp16_0, tensor var_4689_cast_fp16_1 = split(axis = var_4689_axis_0, split_sizes = var_4689_split_sizes_0, x = linear_338_cast_fp16)[name = tensor("op_4689_cast_fp16")]; + tensor var_4691_mode_0 = const()[name = tensor("op_4691_mode_0"), val = tensor("EXACT")]; + tensor var_4691_cast_fp16 = gelu(mode = var_4691_mode_0, x = var_4689_cast_fp16_1)[name = tensor("op_4691_cast_fp16")]; + tensor input_809_cast_fp16 = mul(x = var_4689_cast_fp16_0, y = var_4691_cast_fp16)[name = tensor("input_809_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(852356672))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(857271936))), name = tensor("mid_block_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(857272128)))]; + tensor linear_339_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_7_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_7_ff_net_2_weight_to_fp16_palettized, x = input_809_cast_fp16)[name = tensor("linear_339_cast_fp16")]; + tensor hidden_states_709_cast_fp16 = add(x = linear_339_cast_fp16, y = input_805_cast_fp16)[name = tensor("hidden_states_709_cast_fp16")]; + tensor hidden_states_711_axes_0 = const()[name = tensor("hidden_states_711_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_8_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(857274752)))]; + tensor mid_block_attentions_0_transformer_blocks_8_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(857277376)))]; + tensor hidden_states_711_cast_fp16 = layer_norm(axes = hidden_states_711_axes_0, beta = mid_block_attentions_0_transformer_blocks_8_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_8_norm1_weight_to_fp16, x = hidden_states_709_cast_fp16)[name = tensor("hidden_states_711_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(857280000))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(858508864))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_340_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_711_cast_fp16)[name = tensor("linear_340_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(858509056))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(859737920))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_341_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_711_cast_fp16)[name = tensor("linear_341_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(859738112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(860966976))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_342_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_711_cast_fp16)[name = tensor("linear_342_cast_fp16")]; + tensor var_4726 = const()[name = tensor("op_4726"), val = tensor([1, -1, 20, 64])]; + tensor var_4727_cast_fp16 = reshape(shape = var_4726, x = linear_340_cast_fp16)[name = tensor("op_4727_cast_fp16")]; + tensor var_4729 = const()[name = tensor("op_4729"), val = tensor([1, -1, 20, 64])]; + tensor var_4730_cast_fp16 = reshape(shape = var_4729, x = linear_341_cast_fp16)[name = tensor("op_4730_cast_fp16")]; + tensor var_4732 = const()[name = tensor("op_4732"), val = tensor([1, -1, 20, 64])]; + tensor var_4733_cast_fp16 = reshape(shape = var_4732, x = linear_342_cast_fp16)[name = tensor("op_4733_cast_fp16")]; + tensor value_259_perm_0 = const()[name = tensor("value_259_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_83_y_0_to_fp16 = const()[name = tensor("mul_83_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_83_cast_fp16 = mul(x = var_4727_cast_fp16, y = mul_83_y_0_to_fp16)[name = tensor("mul_83_cast_fp16")]; + tensor matmul_64_transpose_y_0 = const()[name = tensor("matmul_64_transpose_y_0"), val = tensor(true)]; + tensor matmul_64_transpose_x_0 = const()[name = tensor("matmul_64_transpose_x_0"), val = tensor(false)]; + tensor transpose_400_perm_0 = const()[name = tensor("transpose_400_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_401_perm_0 = const()[name = tensor("transpose_401_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_423 = transpose(perm = transpose_401_perm_0, x = var_4730_cast_fp16)[name = tensor("transpose_423")]; + tensor transpose_424 = transpose(perm = transpose_400_perm_0, x = mul_83_cast_fp16)[name = tensor("transpose_424")]; + tensor matmul_64_cast_fp16 = matmul(transpose_x = matmul_64_transpose_x_0, transpose_y = matmul_64_transpose_y_0, x = transpose_424, y = transpose_423)[name = tensor("matmul_64_cast_fp16")]; + tensor softmax_64_axis_0 = const()[name = tensor("softmax_64_axis_0"), val = tensor(-1)]; + tensor softmax_64_cast_fp16 = softmax(axis = softmax_64_axis_0, x = matmul_64_cast_fp16)[name = tensor("softmax_64_cast_fp16")]; + tensor hidden_states_713_transpose_x_0 = const()[name = tensor("hidden_states_713_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_713_transpose_y_0 = const()[name = tensor("hidden_states_713_transpose_y_0"), val = tensor(false)]; + tensor transpose_422 = transpose(perm = value_259_perm_0, x = var_4733_cast_fp16)[name = tensor("transpose_422")]; + tensor hidden_states_713_cast_fp16 = matmul(transpose_x = hidden_states_713_transpose_x_0, transpose_y = hidden_states_713_transpose_y_0, x = softmax_64_cast_fp16, y = transpose_422)[name = tensor("hidden_states_713_cast_fp16")]; + tensor var_4736_perm_0 = const()[name = tensor("op_4736_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4740 = const()[name = tensor("op_4740"), val = tensor([1, -1, 1280])]; + tensor transpose_421 = transpose(perm = var_4736_perm_0, x = hidden_states_713_cast_fp16)[name = tensor("transpose_421")]; + tensor hidden_states_715_cast_fp16 = reshape(shape = var_4740, x = transpose_421)[name = tensor("hidden_states_715_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(860967168))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(862196032))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(862196224)))]; + tensor linear_343_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_715_cast_fp16)[name = tensor("linear_343_cast_fp16")]; + tensor input_817_cast_fp16 = add(x = linear_343_cast_fp16, y = hidden_states_709_cast_fp16)[name = tensor("input_817_cast_fp16")]; + tensor input_819_axes_0 = const()[name = tensor("input_819_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_8_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(862198848)))]; + tensor mid_block_attentions_0_transformer_blocks_8_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(862201472)))]; + tensor input_819_cast_fp16 = layer_norm(axes = input_819_axes_0, beta = mid_block_attentions_0_transformer_blocks_8_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_8_norm2_weight_to_fp16, x = input_817_cast_fp16)[name = tensor("input_819_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(862204096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(863432960))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_344_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn2_to_q_weight_to_fp16_palettized, x = input_819_cast_fp16)[name = tensor("linear_344_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(863433152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(865399296))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_345_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_345_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(865399488))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(867365632))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_346_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_346_cast_fp16")]; + tensor var_4772 = const()[name = tensor("op_4772"), val = tensor([1, -1, 20, 64])]; + tensor var_4773_cast_fp16 = reshape(shape = var_4772, x = linear_344_cast_fp16)[name = tensor("op_4773_cast_fp16")]; + tensor var_4775 = const()[name = tensor("op_4775"), val = tensor([1, -1, 20, 64])]; + tensor var_4776_cast_fp16 = reshape(shape = var_4775, x = linear_345_cast_fp16)[name = tensor("op_4776_cast_fp16")]; + tensor var_4778 = const()[name = tensor("op_4778"), val = tensor([1, -1, 20, 64])]; + tensor var_4779_cast_fp16 = reshape(shape = var_4778, x = linear_346_cast_fp16)[name = tensor("op_4779_cast_fp16")]; + tensor value_263_perm_0 = const()[name = tensor("value_263_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_84_y_0_to_fp16 = const()[name = tensor("mul_84_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_84_cast_fp16 = mul(x = var_4773_cast_fp16, y = mul_84_y_0_to_fp16)[name = tensor("mul_84_cast_fp16")]; + tensor matmul_65_transpose_y_0 = const()[name = tensor("matmul_65_transpose_y_0"), val = tensor(true)]; + tensor matmul_65_transpose_x_0 = const()[name = tensor("matmul_65_transpose_x_0"), val = tensor(false)]; + tensor transpose_402_perm_0 = const()[name = tensor("transpose_402_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_403_perm_0 = const()[name = tensor("transpose_403_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_419 = transpose(perm = transpose_403_perm_0, x = var_4776_cast_fp16)[name = tensor("transpose_419")]; + tensor transpose_420 = transpose(perm = transpose_402_perm_0, x = mul_84_cast_fp16)[name = tensor("transpose_420")]; + tensor matmul_65_cast_fp16 = matmul(transpose_x = matmul_65_transpose_x_0, transpose_y = matmul_65_transpose_y_0, x = transpose_420, y = transpose_419)[name = tensor("matmul_65_cast_fp16")]; + tensor softmax_65_axis_0 = const()[name = tensor("softmax_65_axis_0"), val = tensor(-1)]; + tensor softmax_65_cast_fp16 = softmax(axis = softmax_65_axis_0, x = matmul_65_cast_fp16)[name = tensor("softmax_65_cast_fp16")]; + tensor hidden_states_719_transpose_x_0 = const()[name = tensor("hidden_states_719_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_719_transpose_y_0 = const()[name = tensor("hidden_states_719_transpose_y_0"), val = tensor(false)]; + tensor transpose_418 = transpose(perm = value_263_perm_0, x = var_4779_cast_fp16)[name = tensor("transpose_418")]; + tensor hidden_states_719_cast_fp16 = matmul(transpose_x = hidden_states_719_transpose_x_0, transpose_y = hidden_states_719_transpose_y_0, x = softmax_65_cast_fp16, y = transpose_418)[name = tensor("hidden_states_719_cast_fp16")]; + tensor var_4782_perm_0 = const()[name = tensor("op_4782_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4786 = const()[name = tensor("op_4786"), val = tensor([1, -1, 1280])]; + tensor transpose_417 = transpose(perm = var_4782_perm_0, x = hidden_states_719_cast_fp16)[name = tensor("transpose_417")]; + tensor hidden_states_721_cast_fp16 = reshape(shape = var_4786, x = transpose_417)[name = tensor("hidden_states_721_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(867365824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(868594688))), name = tensor("mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(868594880)))]; + tensor linear_347_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_721_cast_fp16)[name = tensor("linear_347_cast_fp16")]; + tensor input_825_cast_fp16 = add(x = linear_347_cast_fp16, y = input_817_cast_fp16)[name = tensor("input_825_cast_fp16")]; + tensor input_827_axes_0 = const()[name = tensor("input_827_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_8_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(868597504)))]; + tensor mid_block_attentions_0_transformer_blocks_8_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(868600128)))]; + tensor input_827_cast_fp16 = layer_norm(axes = input_827_axes_0, beta = mid_block_attentions_0_transformer_blocks_8_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_8_norm3_weight_to_fp16, x = input_825_cast_fp16)[name = tensor("input_827_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(868602752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(878433216))), name = tensor("mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(878433408))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(878441152))), name = tensor("mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_348_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_8_ff_net_0_proj_weight_to_fp16_palettized, x = input_827_cast_fp16)[name = tensor("linear_348_cast_fp16")]; + tensor var_4808_split_sizes_0 = const()[name = tensor("op_4808_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4808_axis_0 = const()[name = tensor("op_4808_axis_0"), val = tensor(-1)]; + tensor var_4808_cast_fp16_0, tensor var_4808_cast_fp16_1 = split(axis = var_4808_axis_0, split_sizes = var_4808_split_sizes_0, x = linear_348_cast_fp16)[name = tensor("op_4808_cast_fp16")]; + tensor var_4810_mode_0 = const()[name = tensor("op_4810_mode_0"), val = tensor("EXACT")]; + tensor var_4810_cast_fp16 = gelu(mode = var_4810_mode_0, x = var_4808_cast_fp16_1)[name = tensor("op_4810_cast_fp16")]; + tensor input_829_cast_fp16 = mul(x = var_4808_cast_fp16_0, y = var_4810_cast_fp16)[name = tensor("input_829_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(878441344))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(883356608))), name = tensor("mid_block_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(883356800)))]; + tensor linear_349_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_8_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_8_ff_net_2_weight_to_fp16_palettized, x = input_829_cast_fp16)[name = tensor("linear_349_cast_fp16")]; + tensor hidden_states_729_cast_fp16 = add(x = linear_349_cast_fp16, y = input_825_cast_fp16)[name = tensor("hidden_states_729_cast_fp16")]; + tensor hidden_states_731_axes_0 = const()[name = tensor("hidden_states_731_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_9_norm1_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm1_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(883359424)))]; + tensor mid_block_attentions_0_transformer_blocks_9_norm1_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(883362048)))]; + tensor hidden_states_731_cast_fp16 = layer_norm(axes = hidden_states_731_axes_0, beta = mid_block_attentions_0_transformer_blocks_9_norm1_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_9_norm1_weight_to_fp16, x = hidden_states_729_cast_fp16)[name = tensor("hidden_states_731_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(883364672))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(884593536))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_350_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn1_to_q_weight_to_fp16_palettized, x = hidden_states_731_cast_fp16)[name = tensor("linear_350_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(884593728))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(885822592))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_351_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn1_to_k_weight_to_fp16_palettized, x = hidden_states_731_cast_fp16)[name = tensor("linear_351_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(885822784))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(887051648))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_352_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn1_to_v_weight_to_fp16_palettized, x = hidden_states_731_cast_fp16)[name = tensor("linear_352_cast_fp16")]; + tensor var_4845 = const()[name = tensor("op_4845"), val = tensor([1, -1, 20, 64])]; + tensor var_4846_cast_fp16 = reshape(shape = var_4845, x = linear_350_cast_fp16)[name = tensor("op_4846_cast_fp16")]; + tensor var_4848 = const()[name = tensor("op_4848"), val = tensor([1, -1, 20, 64])]; + tensor var_4849_cast_fp16 = reshape(shape = var_4848, x = linear_351_cast_fp16)[name = tensor("op_4849_cast_fp16")]; + tensor var_4851 = const()[name = tensor("op_4851"), val = tensor([1, -1, 20, 64])]; + tensor var_4852_cast_fp16 = reshape(shape = var_4851, x = linear_352_cast_fp16)[name = tensor("op_4852_cast_fp16")]; + tensor value_267_perm_0 = const()[name = tensor("value_267_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_85_y_0_to_fp16 = const()[name = tensor("mul_85_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_85_cast_fp16 = mul(x = var_4846_cast_fp16, y = mul_85_y_0_to_fp16)[name = tensor("mul_85_cast_fp16")]; + tensor matmul_66_transpose_y_0 = const()[name = tensor("matmul_66_transpose_y_0"), val = tensor(true)]; + tensor matmul_66_transpose_x_0 = const()[name = tensor("matmul_66_transpose_x_0"), val = tensor(false)]; + tensor transpose_404_perm_0 = const()[name = tensor("transpose_404_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_405_perm_0 = const()[name = tensor("transpose_405_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_415 = transpose(perm = transpose_405_perm_0, x = var_4849_cast_fp16)[name = tensor("transpose_415")]; + tensor transpose_416 = transpose(perm = transpose_404_perm_0, x = mul_85_cast_fp16)[name = tensor("transpose_416")]; + tensor matmul_66_cast_fp16 = matmul(transpose_x = matmul_66_transpose_x_0, transpose_y = matmul_66_transpose_y_0, x = transpose_416, y = transpose_415)[name = tensor("matmul_66_cast_fp16")]; + tensor softmax_66_axis_0 = const()[name = tensor("softmax_66_axis_0"), val = tensor(-1)]; + tensor softmax_66_cast_fp16 = softmax(axis = softmax_66_axis_0, x = matmul_66_cast_fp16)[name = tensor("softmax_66_cast_fp16")]; + tensor hidden_states_733_transpose_x_0 = const()[name = tensor("hidden_states_733_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_733_transpose_y_0 = const()[name = tensor("hidden_states_733_transpose_y_0"), val = tensor(false)]; + tensor transpose_414 = transpose(perm = value_267_perm_0, x = var_4852_cast_fp16)[name = tensor("transpose_414")]; + tensor hidden_states_733_cast_fp16 = matmul(transpose_x = hidden_states_733_transpose_x_0, transpose_y = hidden_states_733_transpose_y_0, x = softmax_66_cast_fp16, y = transpose_414)[name = tensor("hidden_states_733_cast_fp16")]; + tensor var_4855_perm_0 = const()[name = tensor("op_4855_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4859 = const()[name = tensor("op_4859"), val = tensor([1, -1, 1280])]; + tensor transpose_413 = transpose(perm = var_4855_perm_0, x = hidden_states_733_cast_fp16)[name = tensor("transpose_413")]; + tensor hidden_states_735_cast_fp16 = reshape(shape = var_4859, x = transpose_413)[name = tensor("hidden_states_735_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(887051840))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(888280704))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(888280896)))]; + tensor linear_353_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn1_to_out_0_weight_to_fp16_palettized, x = hidden_states_735_cast_fp16)[name = tensor("linear_353_cast_fp16")]; + tensor input_837_cast_fp16 = add(x = linear_353_cast_fp16, y = hidden_states_729_cast_fp16)[name = tensor("input_837_cast_fp16")]; + tensor input_839_axes_0 = const()[name = tensor("input_839_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_9_norm2_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm2_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(888283520)))]; + tensor mid_block_attentions_0_transformer_blocks_9_norm2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(888286144)))]; + tensor input_839_cast_fp16 = layer_norm(axes = input_839_axes_0, beta = mid_block_attentions_0_transformer_blocks_9_norm2_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_9_norm2_weight_to_fp16, x = input_837_cast_fp16)[name = tensor("input_839_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(888288768))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(889517632))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor linear_354_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn2_to_q_weight_to_fp16_palettized, x = input_839_cast_fp16)[name = tensor("linear_354_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(889517824))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(891483968))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_355_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn2_to_k_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_355_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(891484160))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(893450304))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized"), shape = tensor([1280, 2048])]; + tensor linear_356_cast_fp16 = linear(bias = add_23_mean_0_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn2_to_v_weight_to_fp16_palettized, x = encoder_hidden_states)[name = tensor("linear_356_cast_fp16")]; + tensor var_4891 = const()[name = tensor("op_4891"), val = tensor([1, -1, 20, 64])]; + tensor var_4892_cast_fp16 = reshape(shape = var_4891, x = linear_354_cast_fp16)[name = tensor("op_4892_cast_fp16")]; + tensor var_4894 = const()[name = tensor("op_4894"), val = tensor([1, -1, 20, 64])]; + tensor var_4895_cast_fp16 = reshape(shape = var_4894, x = linear_355_cast_fp16)[name = tensor("op_4895_cast_fp16")]; + tensor var_4897 = const()[name = tensor("op_4897"), val = tensor([1, -1, 20, 64])]; + tensor var_4898_cast_fp16 = reshape(shape = var_4897, x = linear_356_cast_fp16)[name = tensor("op_4898_cast_fp16")]; + tensor value_perm_0 = const()[name = tensor("value_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor mul_86_y_0_to_fp16 = const()[name = tensor("mul_86_y_0_to_fp16"), val = tensor(0x1p-3)]; + tensor mul_86_cast_fp16 = mul(x = var_4892_cast_fp16, y = mul_86_y_0_to_fp16)[name = tensor("mul_86_cast_fp16")]; + tensor matmul_67_transpose_y_0 = const()[name = tensor("matmul_67_transpose_y_0"), val = tensor(true)]; + tensor matmul_67_transpose_x_0 = const()[name = tensor("matmul_67_transpose_x_0"), val = tensor(false)]; + tensor transpose_406_perm_0 = const()[name = tensor("transpose_406_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_407_perm_0 = const()[name = tensor("transpose_407_perm_0"), val = tensor([0, 2, -3, -1])]; + tensor transpose_411 = transpose(perm = transpose_407_perm_0, x = var_4895_cast_fp16)[name = tensor("transpose_411")]; + tensor transpose_412 = transpose(perm = transpose_406_perm_0, x = mul_86_cast_fp16)[name = tensor("transpose_412")]; + tensor matmul_67_cast_fp16 = matmul(transpose_x = matmul_67_transpose_x_0, transpose_y = matmul_67_transpose_y_0, x = transpose_412, y = transpose_411)[name = tensor("matmul_67_cast_fp16")]; + tensor softmax_67_axis_0 = const()[name = tensor("softmax_67_axis_0"), val = tensor(-1)]; + tensor softmax_67_cast_fp16 = softmax(axis = softmax_67_axis_0, x = matmul_67_cast_fp16)[name = tensor("softmax_67_cast_fp16")]; + tensor hidden_states_739_transpose_x_0 = const()[name = tensor("hidden_states_739_transpose_x_0"), val = tensor(false)]; + tensor hidden_states_739_transpose_y_0 = const()[name = tensor("hidden_states_739_transpose_y_0"), val = tensor(false)]; + tensor transpose_410 = transpose(perm = value_perm_0, x = var_4898_cast_fp16)[name = tensor("transpose_410")]; + tensor hidden_states_739_cast_fp16 = matmul(transpose_x = hidden_states_739_transpose_x_0, transpose_y = hidden_states_739_transpose_y_0, x = softmax_67_cast_fp16, y = transpose_410)[name = tensor("hidden_states_739_cast_fp16")]; + tensor var_4901_perm_0 = const()[name = tensor("op_4901_perm_0"), val = tensor([0, 2, 1, 3])]; + tensor var_4905 = const()[name = tensor("op_4905"), val = tensor([1, -1, 1280])]; + tensor transpose_409 = transpose(perm = var_4901_perm_0, x = hidden_states_739_cast_fp16)[name = tensor("transpose_409")]; + tensor hidden_states_741_cast_fp16 = reshape(shape = var_4905, x = transpose_409)[name = tensor("hidden_states_741_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(893450496))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(894679360))), name = tensor("mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(894679552)))]; + tensor linear_357_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_attn2_to_out_0_weight_to_fp16_palettized, x = hidden_states_741_cast_fp16)[name = tensor("linear_357_cast_fp16")]; + tensor input_845_cast_fp16 = add(x = linear_357_cast_fp16, y = input_837_cast_fp16)[name = tensor("input_845_cast_fp16")]; + tensor input_847_axes_0 = const()[name = tensor("input_847_axes_0"), val = tensor([-1])]; + tensor mid_block_attentions_0_transformer_blocks_9_norm3_weight_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm3_weight_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(894682176)))]; + tensor mid_block_attentions_0_transformer_blocks_9_norm3_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_norm3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(894684800)))]; + tensor input_847_cast_fp16 = layer_norm(axes = input_847_axes_0, beta = mid_block_attentions_0_transformer_blocks_9_norm3_bias_to_fp16, epsilon = var_3647_to_fp16, gamma = mid_block_attentions_0_transformer_blocks_9_norm3_weight_to_fp16, x = input_845_cast_fp16)[name = tensor("input_847_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(894687424))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(904517888))), name = tensor("mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized"), shape = tensor([10240, 1280])]; + tensor mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(904518080))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(904525824))), name = tensor("mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized"), shape = tensor([10240])]; + tensor linear_358_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_bias_to_fp16_palettized, weight = mid_block_attentions_0_transformer_blocks_9_ff_net_0_proj_weight_to_fp16_palettized, x = input_847_cast_fp16)[name = tensor("linear_358_cast_fp16")]; + tensor var_4927_split_sizes_0 = const()[name = tensor("op_4927_split_sizes_0"), val = tensor([5120, 5120])]; + tensor var_4927_axis_0 = const()[name = tensor("op_4927_axis_0"), val = tensor(-1)]; + tensor var_4927_cast_fp16_0, tensor var_4927_cast_fp16_1 = split(axis = var_4927_axis_0, split_sizes = var_4927_split_sizes_0, x = linear_358_cast_fp16)[name = tensor("op_4927_cast_fp16")]; + tensor var_4929_mode_0 = const()[name = tensor("op_4929_mode_0"), val = tensor("EXACT")]; + tensor var_4929_cast_fp16 = gelu(mode = var_4929_mode_0, x = var_4927_cast_fp16_1)[name = tensor("op_4929_cast_fp16")]; + tensor input_849_cast_fp16 = mul(x = var_4927_cast_fp16_0, y = var_4929_cast_fp16)[name = tensor("input_849_cast_fp16")]; + tensor mid_block_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(904526016))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(909441280))), name = tensor("mid_block_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized"), shape = tensor([1280, 5120])]; + tensor mid_block_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(909441472)))]; + tensor linear_359_cast_fp16 = linear(bias = mid_block_attentions_0_transformer_blocks_9_ff_net_2_bias_to_fp16, weight = mid_block_attentions_0_transformer_blocks_9_ff_net_2_weight_to_fp16_palettized, x = input_849_cast_fp16)[name = tensor("linear_359_cast_fp16")]; + tensor input_853_cast_fp16 = add(x = linear_359_cast_fp16, y = input_845_cast_fp16)[name = tensor("input_853_cast_fp16")]; + tensor mid_block_attentions_0_proj_out_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(909444096))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(910672960))), name = tensor("mid_block_attentions_0_proj_out_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_attentions_0_proj_out_bias_to_fp16 = const()[name = tensor("mid_block_attentions_0_proj_out_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(910673152)))]; + tensor linear_360_cast_fp16 = linear(bias = mid_block_attentions_0_proj_out_bias_to_fp16, weight = mid_block_attentions_0_proj_out_weight_to_fp16_palettized, x = input_853_cast_fp16)[name = tensor("linear_360_cast_fp16")]; + tensor var_4939 = const()[name = tensor("op_4939"), val = tensor([1, 32, 32, 1280])]; + tensor var_4940_cast_fp16 = reshape(shape = var_4939, x = linear_360_cast_fp16)[name = tensor("op_4940_cast_fp16")]; + tensor var_4941 = const()[name = tensor("op_4941"), val = tensor([0, 3, 1, 2])]; + tensor transpose_408 = transpose(perm = var_4941, x = var_4940_cast_fp16)[name = tensor("transpose_408")]; + tensor input_855_cast_fp16 = add(x = transpose_408, y = var_3699_cast_fp16)[name = tensor("input_855_cast_fp16")]; + tensor reshape_76_shape_0 = const()[name = tensor("reshape_76_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_76_cast_fp16 = reshape(shape = reshape_76_shape_0, x = input_855_cast_fp16)[name = tensor("reshape_76_cast_fp16")]; + tensor reduce_mean_57_axes_0 = const()[name = tensor("reduce_mean_57_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_57_keep_dims_0 = const()[name = tensor("reduce_mean_57_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_57_cast_fp16 = reduce_mean(axes = reduce_mean_57_axes_0, keep_dims = reduce_mean_57_keep_dims_0, x = reshape_76_cast_fp16)[name = tensor("reduce_mean_57_cast_fp16")]; + tensor sub_38_cast_fp16 = sub(x = reshape_76_cast_fp16, y = reduce_mean_57_cast_fp16)[name = tensor("sub_38_cast_fp16")]; + tensor square_19_cast_fp16 = square(x = sub_38_cast_fp16)[name = tensor("square_19_cast_fp16")]; + tensor reduce_mean_59_axes_0 = const()[name = tensor("reduce_mean_59_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_59_keep_dims_0 = const()[name = tensor("reduce_mean_59_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_59_cast_fp16 = reduce_mean(axes = reduce_mean_59_axes_0, keep_dims = reduce_mean_59_keep_dims_0, x = square_19_cast_fp16)[name = tensor("reduce_mean_59_cast_fp16")]; + tensor add_38_y_0_to_fp16 = const()[name = tensor("add_38_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_38_cast_fp16 = add(x = reduce_mean_59_cast_fp16, y = add_38_y_0_to_fp16)[name = tensor("add_38_cast_fp16")]; + tensor sqrt_19_cast_fp16 = sqrt(x = add_38_cast_fp16)[name = tensor("sqrt_19_cast_fp16")]; + tensor real_div_19_cast_fp16 = real_div(x = sub_38_cast_fp16, y = sqrt_19_cast_fp16)[name = tensor("real_div_19_cast_fp16")]; + tensor reshape_77_shape_0 = const()[name = tensor("reshape_77_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_77_cast_fp16 = reshape(shape = reshape_77_shape_0, x = real_div_19_cast_fp16)[name = tensor("reshape_77_cast_fp16")]; + tensor add_39_gamma_0_to_fp16 = const()[name = tensor("add_39_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(910675776)))]; + tensor add_39_beta_0_to_fp16 = const()[name = tensor("add_39_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(910678400)))]; + tensor add_39_epsilon_0_to_fp16 = const()[name = tensor("add_39_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_39_cast_fp16 = batch_norm(beta = add_39_beta_0_to_fp16, epsilon = add_39_epsilon_0_to_fp16, gamma = add_39_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_77_cast_fp16)[name = tensor("add_39_cast_fp16")]; + tensor input_859_cast_fp16 = silu(x = add_39_cast_fp16)[name = tensor("input_859_cast_fp16")]; + tensor var_4956 = const()[name = tensor("op_4956"), val = tensor([1, 1])]; + tensor var_4958 = const()[name = tensor("op_4958"), val = tensor([1, 1])]; + tensor hidden_states_753_pad_type_0 = const()[name = tensor("hidden_states_753_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_753_pad_0 = const()[name = tensor("hidden_states_753_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor mid_block_resnets_1_conv1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(910681024))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(921740288))), name = tensor("mid_block_resnets_1_conv1_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor mid_block_resnets_1_conv1_bias_to_fp16 = const()[name = tensor("mid_block_resnets_1_conv1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(921740480)))]; + tensor hidden_states_753_cast_fp16 = conv(bias = mid_block_resnets_1_conv1_bias_to_fp16, dilations = var_4958, groups = var_3651, pad = hidden_states_753_pad_0, pad_type = hidden_states_753_pad_type_0, strides = var_4956, weight = mid_block_resnets_1_conv1_weight_to_fp16_palettized, x = input_859_cast_fp16)[name = tensor("hidden_states_753_cast_fp16")]; + tensor mid_block_resnets_1_time_emb_proj_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(921743104))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922971968))), name = tensor("mid_block_resnets_1_time_emb_proj_weight_to_fp16_palettized"), shape = tensor([1280, 1280])]; + tensor mid_block_resnets_1_time_emb_proj_bias_to_fp16 = const()[name = tensor("mid_block_resnets_1_time_emb_proj_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922972160)))]; + tensor linear_361_cast_fp16 = linear(bias = mid_block_resnets_1_time_emb_proj_bias_to_fp16, weight = mid_block_resnets_1_time_emb_proj_weight_to_fp16_palettized, x = input_49_cast_fp16)[name = tensor("linear_361_cast_fp16")]; + tensor var_4967_axes_0 = const()[name = tensor("op_4967_axes_0"), val = tensor([2])]; + tensor var_4967_cast_fp16 = expand_dims(axes = var_4967_axes_0, x = linear_361_cast_fp16)[name = tensor("op_4967_cast_fp16")]; + tensor temb_axes_0 = const()[name = tensor("temb_axes_0"), val = tensor([3])]; + tensor temb_cast_fp16 = expand_dims(axes = temb_axes_0, x = var_4967_cast_fp16)[name = tensor("temb_cast_fp16")]; + tensor input_863_cast_fp16 = add(x = hidden_states_753_cast_fp16, y = temb_cast_fp16)[name = tensor("input_863_cast_fp16")]; + tensor reshape_80_shape_0 = const()[name = tensor("reshape_80_shape_0"), val = tensor([1, 32, 40, 32, 32])]; + tensor reshape_80_cast_fp16 = reshape(shape = reshape_80_shape_0, x = input_863_cast_fp16)[name = tensor("reshape_80_cast_fp16")]; + tensor reduce_mean_60_axes_0 = const()[name = tensor("reduce_mean_60_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_60_keep_dims_0 = const()[name = tensor("reduce_mean_60_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_60_cast_fp16 = reduce_mean(axes = reduce_mean_60_axes_0, keep_dims = reduce_mean_60_keep_dims_0, x = reshape_80_cast_fp16)[name = tensor("reduce_mean_60_cast_fp16")]; + tensor sub_40_cast_fp16 = sub(x = reshape_80_cast_fp16, y = reduce_mean_60_cast_fp16)[name = tensor("sub_40_cast_fp16")]; + tensor square_20_cast_fp16 = square(x = sub_40_cast_fp16)[name = tensor("square_20_cast_fp16")]; + tensor reduce_mean_62_axes_0 = const()[name = tensor("reduce_mean_62_axes_0"), val = tensor([2, 3, 4])]; + tensor reduce_mean_62_keep_dims_0 = const()[name = tensor("reduce_mean_62_keep_dims_0"), val = tensor(true)]; + tensor reduce_mean_62_cast_fp16 = reduce_mean(axes = reduce_mean_62_axes_0, keep_dims = reduce_mean_62_keep_dims_0, x = square_20_cast_fp16)[name = tensor("reduce_mean_62_cast_fp16")]; + tensor add_40_y_0_to_fp16 = const()[name = tensor("add_40_y_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_40_cast_fp16 = add(x = reduce_mean_62_cast_fp16, y = add_40_y_0_to_fp16)[name = tensor("add_40_cast_fp16")]; + tensor sqrt_20_cast_fp16 = sqrt(x = add_40_cast_fp16)[name = tensor("sqrt_20_cast_fp16")]; + tensor real_div_20_cast_fp16 = real_div(x = sub_40_cast_fp16, y = sqrt_20_cast_fp16)[name = tensor("real_div_20_cast_fp16")]; + tensor reshape_81_shape_0 = const()[name = tensor("reshape_81_shape_0"), val = tensor([1, 1280, 32, 32])]; + tensor reshape_81_cast_fp16 = reshape(shape = reshape_81_shape_0, x = real_div_20_cast_fp16)[name = tensor("reshape_81_cast_fp16")]; + tensor add_41_gamma_0_to_fp16 = const()[name = tensor("add_41_gamma_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922974784)))]; + tensor add_41_beta_0_to_fp16 = const()[name = tensor("add_41_beta_0_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922977408)))]; + tensor add_41_epsilon_0_to_fp16 = const()[name = tensor("add_41_epsilon_0_to_fp16"), val = tensor(0x1.5p-17)]; + tensor add_41_cast_fp16 = batch_norm(beta = add_41_beta_0_to_fp16, epsilon = add_41_epsilon_0_to_fp16, gamma = add_41_gamma_0_to_fp16, mean = add_23_mean_0_to_fp16, variance = add_23_variance_0_to_fp16, x = reshape_81_cast_fp16)[name = tensor("add_41_cast_fp16")]; + tensor input_867_cast_fp16 = silu(x = add_41_cast_fp16)[name = tensor("input_867_cast_fp16")]; + tensor var_4977 = const()[name = tensor("op_4977"), val = tensor([1, 1])]; + tensor var_4979 = const()[name = tensor("op_4979"), val = tensor([1, 1])]; + tensor hidden_states_pad_type_0 = const()[name = tensor("hidden_states_pad_type_0"), val = tensor("custom")]; + tensor hidden_states_pad_0 = const()[name = tensor("hidden_states_pad_0"), val = tensor([1, 1, 1, 1])]; + tensor mid_block_resnets_1_conv2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(922980032))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934039296))), name = tensor("mid_block_resnets_1_conv2_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 3, 3])]; + tensor mid_block_resnets_1_conv2_bias_to_fp16 = const()[name = tensor("mid_block_resnets_1_conv2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934039488)))]; + tensor hidden_states_cast_fp16 = conv(bias = mid_block_resnets_1_conv2_bias_to_fp16, dilations = var_4979, groups = var_3651, pad = hidden_states_pad_0, pad_type = hidden_states_pad_type_0, strides = var_4977, weight = mid_block_resnets_1_conv2_weight_to_fp16_palettized, x = input_867_cast_fp16)[name = tensor("hidden_states_cast_fp16")]; + tensor var_4982_cast_fp16 = add(x = input_855_cast_fp16, y = hidden_states_cast_fp16)[name = tensor("op_4982_cast_fp16")]; + tensor var_4988 = const()[name = tensor("op_4988"), val = tensor(1)]; + tensor var_4991 = const()[name = tensor("op_4991"), val = tensor([1, 1])]; + tensor var_4993 = const()[name = tensor("op_4993"), val = tensor([1, 1])]; + tensor sample_5_pad_type_0 = const()[name = tensor("sample_5_pad_type_0"), val = tensor("custom")]; + tensor sample_5_pad_0 = const()[name = tensor("sample_5_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_0_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934042112))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934118976))), name = tensor("controlnet_down_blocks_0_weight_to_fp16_palettized"), shape = tensor([320, 320, 1, 1])]; + tensor controlnet_down_blocks_0_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_0_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934119168)))]; + tensor sample_5_cast_fp16 = conv(bias = controlnet_down_blocks_0_bias_to_fp16, dilations = var_4993, groups = var_4988, pad = sample_5_pad_0, pad_type = sample_5_pad_type_0, strides = var_4991, weight = controlnet_down_blocks_0_weight_to_fp16_palettized, x = input_41_cast_fp16)[name = tensor("sample_5_cast_fp16")]; + tensor var_4999 = const()[name = tensor("op_4999"), val = tensor(1)]; + tensor var_5002 = const()[name = tensor("op_5002"), val = tensor([1, 1])]; + tensor var_5004 = const()[name = tensor("op_5004"), val = tensor([1, 1])]; + tensor sample_7_pad_type_0 = const()[name = tensor("sample_7_pad_type_0"), val = tensor("custom")]; + tensor sample_7_pad_0 = const()[name = tensor("sample_7_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_1_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934119872))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934196736))), name = tensor("controlnet_down_blocks_1_weight_to_fp16_palettized"), shape = tensor([320, 320, 1, 1])]; + tensor controlnet_down_blocks_1_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_1_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934196928)))]; + tensor sample_7_cast_fp16 = conv(bias = controlnet_down_blocks_1_bias_to_fp16, dilations = var_5004, groups = var_4999, pad = sample_7_pad_0, pad_type = sample_7_pad_type_0, strides = var_5002, weight = controlnet_down_blocks_1_weight_to_fp16_palettized, x = var_302_cast_fp16)[name = tensor("sample_7_cast_fp16")]; + tensor var_5010 = const()[name = tensor("op_5010"), val = tensor(1)]; + tensor var_5013 = const()[name = tensor("op_5013"), val = tensor([1, 1])]; + tensor var_5015 = const()[name = tensor("op_5015"), val = tensor([1, 1])]; + tensor sample_9_pad_type_0 = const()[name = tensor("sample_9_pad_type_0"), val = tensor("custom")]; + tensor sample_9_pad_0 = const()[name = tensor("sample_9_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_2_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934197632))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934274496))), name = tensor("controlnet_down_blocks_2_weight_to_fp16_palettized"), shape = tensor([320, 320, 1, 1])]; + tensor controlnet_down_blocks_2_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_2_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934274688)))]; + tensor sample_9_cast_fp16 = conv(bias = controlnet_down_blocks_2_bias_to_fp16, dilations = var_5015, groups = var_5010, pad = sample_9_pad_0, pad_type = sample_9_pad_type_0, strides = var_5013, weight = controlnet_down_blocks_2_weight_to_fp16_palettized, x = var_342_cast_fp16)[name = tensor("sample_9_cast_fp16")]; + tensor var_5021 = const()[name = tensor("op_5021"), val = tensor(1)]; + tensor var_5024 = const()[name = tensor("op_5024"), val = tensor([1, 1])]; + tensor var_5026 = const()[name = tensor("op_5026"), val = tensor([1, 1])]; + tensor sample_11_pad_type_0 = const()[name = tensor("sample_11_pad_type_0"), val = tensor("custom")]; + tensor sample_11_pad_0 = const()[name = tensor("sample_11_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_3_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934275392))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934352256))), name = tensor("controlnet_down_blocks_3_weight_to_fp16_palettized"), shape = tensor([320, 320, 1, 1])]; + tensor controlnet_down_blocks_3_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_3_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934352448)))]; + tensor sample_11_cast_fp16 = conv(bias = controlnet_down_blocks_3_bias_to_fp16, dilations = var_5026, groups = var_5021, pad = sample_11_pad_0, pad_type = sample_11_pad_type_0, strides = var_5024, weight = controlnet_down_blocks_3_weight_to_fp16_palettized, x = input_75_cast_fp16)[name = tensor("sample_11_cast_fp16")]; + tensor var_5032 = const()[name = tensor("op_5032"), val = tensor(1)]; + tensor var_5035 = const()[name = tensor("op_5035"), val = tensor([1, 1])]; + tensor var_5037 = const()[name = tensor("op_5037"), val = tensor([1, 1])]; + tensor sample_13_pad_type_0 = const()[name = tensor("sample_13_pad_type_0"), val = tensor("custom")]; + tensor sample_13_pad_0 = const()[name = tensor("sample_13_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_4_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934353152))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934660416))), name = tensor("controlnet_down_blocks_4_weight_to_fp16_palettized"), shape = tensor([640, 640, 1, 1])]; + tensor controlnet_down_blocks_4_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_4_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934660608)))]; + tensor sample_13_cast_fp16 = conv(bias = controlnet_down_blocks_4_bias_to_fp16, dilations = var_5037, groups = var_5032, pad = sample_13_pad_0, pad_type = sample_13_pad_type_0, strides = var_5035, weight = controlnet_down_blocks_4_weight_to_fp16_palettized, x = input_135_cast_fp16)[name = tensor("sample_13_cast_fp16")]; + tensor var_5043 = const()[name = tensor("op_5043"), val = tensor(1)]; + tensor var_5046 = const()[name = tensor("op_5046"), val = tensor([1, 1])]; + tensor var_5048 = const()[name = tensor("op_5048"), val = tensor([1, 1])]; + tensor sample_15_pad_type_0 = const()[name = tensor("sample_15_pad_type_0"), val = tensor("custom")]; + tensor sample_15_pad_0 = const()[name = tensor("sample_15_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_5_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934661952))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934969216))), name = tensor("controlnet_down_blocks_5_weight_to_fp16_palettized"), shape = tensor([640, 640, 1, 1])]; + tensor controlnet_down_blocks_5_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_5_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934969408)))]; + tensor sample_15_cast_fp16 = conv(bias = controlnet_down_blocks_5_bias_to_fp16, dilations = var_5048, groups = var_5043, pad = sample_15_pad_0, pad_type = sample_15_pad_type_0, strides = var_5046, weight = controlnet_down_blocks_5_weight_to_fp16_palettized, x = hidden_states_115_cast_fp16)[name = tensor("sample_15_cast_fp16")]; + tensor var_5054 = const()[name = tensor("op_5054"), val = tensor(1)]; + tensor var_5057 = const()[name = tensor("op_5057"), val = tensor([1, 1])]; + tensor var_5059 = const()[name = tensor("op_5059"), val = tensor([1, 1])]; + tensor sample_17_pad_type_0 = const()[name = tensor("sample_17_pad_type_0"), val = tensor("custom")]; + tensor sample_17_pad_0 = const()[name = tensor("sample_17_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_6_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(934970752))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(935278016))), name = tensor("controlnet_down_blocks_6_weight_to_fp16_palettized"), shape = tensor([640, 640, 1, 1])]; + tensor controlnet_down_blocks_6_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_6_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(935278208)))]; + tensor sample_17_cast_fp16 = conv(bias = controlnet_down_blocks_6_bias_to_fp16, dilations = var_5059, groups = var_5054, pad = sample_17_pad_0, pad_type = sample_17_pad_type_0, strides = var_5057, weight = controlnet_down_blocks_6_weight_to_fp16_palettized, x = input_195_cast_fp16)[name = tensor("sample_17_cast_fp16")]; + tensor var_5065 = const()[name = tensor("op_5065"), val = tensor(1)]; + tensor var_5068 = const()[name = tensor("op_5068"), val = tensor([1, 1])]; + tensor var_5070 = const()[name = tensor("op_5070"), val = tensor([1, 1])]; + tensor sample_19_pad_type_0 = const()[name = tensor("sample_19_pad_type_0"), val = tensor("custom")]; + tensor sample_19_pad_0 = const()[name = tensor("sample_19_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_7_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(935279552))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(936508416))), name = tensor("controlnet_down_blocks_7_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 1, 1])]; + tensor controlnet_down_blocks_7_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_7_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(936508608)))]; + tensor sample_19_cast_fp16 = conv(bias = controlnet_down_blocks_7_bias_to_fp16, dilations = var_5070, groups = var_5065, pad = sample_19_pad_0, pad_type = sample_19_pad_type_0, strides = var_5068, weight = controlnet_down_blocks_7_weight_to_fp16_palettized, x = input_415_cast_fp16)[name = tensor("sample_19_cast_fp16")]; + tensor var_5076 = const()[name = tensor("op_5076"), val = tensor(1)]; + tensor var_5079 = const()[name = tensor("op_5079"), val = tensor([1, 1])]; + tensor var_5081 = const()[name = tensor("op_5081"), val = tensor([1, 1])]; + tensor sample_pad_type_0 = const()[name = tensor("sample_pad_type_0"), val = tensor("custom")]; + tensor sample_pad_0 = const()[name = tensor("sample_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_down_blocks_8_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(936511232))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(937740096))), name = tensor("controlnet_down_blocks_8_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 1, 1])]; + tensor controlnet_down_blocks_8_bias_to_fp16 = const()[name = tensor("controlnet_down_blocks_8_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(937740288)))]; + tensor sample_cast_fp16 = conv(bias = controlnet_down_blocks_8_bias_to_fp16, dilations = var_5081, groups = var_5076, pad = sample_pad_0, pad_type = sample_pad_type_0, strides = var_5079, weight = controlnet_down_blocks_8_weight_to_fp16_palettized, x = input_635_cast_fp16)[name = tensor("sample_cast_fp16")]; + tensor var_5087 = const()[name = tensor("op_5087"), val = tensor(1)]; + tensor var_5090 = const()[name = tensor("op_5090"), val = tensor([1, 1])]; + tensor var_5092 = const()[name = tensor("op_5092"), val = tensor([1, 1])]; + tensor mid_block_res_sample_pad_type_0 = const()[name = tensor("mid_block_res_sample_pad_type_0"), val = tensor("custom")]; + tensor mid_block_res_sample_pad_0 = const()[name = tensor("mid_block_res_sample_pad_0"), val = tensor([0, 0, 0, 0])]; + tensor controlnet_mid_block_weight_to_fp16_palettized = constexpr_lut_to_dense()[indices = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(937742912))), lut = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(938971776))), name = tensor("controlnet_mid_block_weight_to_fp16_palettized"), shape = tensor([1280, 1280, 1, 1])]; + tensor controlnet_mid_block_bias_to_fp16 = const()[name = tensor("controlnet_mid_block_bias_to_fp16"), val = tensor(BLOBFILE(path = tensor("@model_path/weights/weight.bin"), offset = tensor(938971968)))]; + tensor mid_block_res_sample_cast_fp16 = conv(bias = controlnet_mid_block_bias_to_fp16, dilations = var_5092, groups = var_5087, pad = mid_block_res_sample_pad_0, pad_type = mid_block_res_sample_pad_type_0, strides = var_5090, weight = controlnet_mid_block_weight_to_fp16_palettized, x = var_4982_cast_fp16)[name = tensor("mid_block_res_sample_cast_fp16")]; + tensor var_5095_cast_fp16 = mul(x = sample_5_cast_fp16, y = conditioning_scale)[name = tensor("op_5095_cast_fp16")]; + tensor var_5095_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5095_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5096_cast_fp16 = mul(x = sample_7_cast_fp16, y = conditioning_scale)[name = tensor("op_5096_cast_fp16")]; + tensor var_5096_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5096_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5097_cast_fp16 = mul(x = sample_9_cast_fp16, y = conditioning_scale)[name = tensor("op_5097_cast_fp16")]; + tensor var_5097_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5097_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5098_cast_fp16 = mul(x = sample_11_cast_fp16, y = conditioning_scale)[name = tensor("op_5098_cast_fp16")]; + tensor var_5098_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5098_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5099_cast_fp16 = mul(x = sample_13_cast_fp16, y = conditioning_scale)[name = tensor("op_5099_cast_fp16")]; + tensor var_5099_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5099_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5100_cast_fp16 = mul(x = sample_15_cast_fp16, y = conditioning_scale)[name = tensor("op_5100_cast_fp16")]; + tensor var_5100_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5100_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5101_cast_fp16 = mul(x = sample_17_cast_fp16, y = conditioning_scale)[name = tensor("op_5101_cast_fp16")]; + tensor var_5101_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5101_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5102_cast_fp16 = mul(x = sample_19_cast_fp16, y = conditioning_scale)[name = tensor("op_5102_cast_fp16")]; + tensor var_5102_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5102_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5103_cast_fp16 = mul(x = sample_cast_fp16, y = conditioning_scale)[name = tensor("op_5103_cast_fp16")]; + tensor var_5103_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5103_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor var_5104_cast_fp16 = mul(x = mid_block_res_sample_cast_fp16, y = conditioning_scale)[name = tensor("op_5104_cast_fp16")]; + tensor var_5104_cast_fp16_to_fp32_dtype_0 = const()[name = tensor("op_5104_cast_fp16_to_fp32_dtype_0"), val = tensor("fp32")]; + tensor additional_residual_9 = cast(dtype = var_5104_cast_fp16_to_fp32_dtype_0, x = var_5104_cast_fp16)[name = tensor("cast_0")]; + tensor additional_residual_8 = cast(dtype = var_5103_cast_fp16_to_fp32_dtype_0, x = var_5103_cast_fp16)[name = tensor("cast_1")]; + tensor additional_residual_7 = cast(dtype = var_5102_cast_fp16_to_fp32_dtype_0, x = var_5102_cast_fp16)[name = tensor("cast_2")]; + tensor additional_residual_6 = cast(dtype = var_5101_cast_fp16_to_fp32_dtype_0, x = var_5101_cast_fp16)[name = tensor("cast_3")]; + tensor additional_residual_5 = cast(dtype = var_5100_cast_fp16_to_fp32_dtype_0, x = var_5100_cast_fp16)[name = tensor("cast_4")]; + tensor additional_residual_4 = cast(dtype = var_5099_cast_fp16_to_fp32_dtype_0, x = var_5099_cast_fp16)[name = tensor("cast_5")]; + tensor additional_residual_3 = cast(dtype = var_5098_cast_fp16_to_fp32_dtype_0, x = var_5098_cast_fp16)[name = tensor("cast_6")]; + tensor additional_residual_2 = cast(dtype = var_5097_cast_fp16_to_fp32_dtype_0, x = var_5097_cast_fp16)[name = tensor("cast_7")]; + tensor additional_residual_1 = cast(dtype = var_5096_cast_fp16_to_fp32_dtype_0, x = var_5096_cast_fp16)[name = tensor("cast_8")]; + tensor additional_residual_0 = cast(dtype = var_5095_cast_fp16_to_fp32_dtype_0, x = var_5095_cast_fp16)[name = tensor("cast_9")]; + } -> (additional_residual_0, additional_residual_1, additional_residual_2, additional_residual_3, additional_residual_4, additional_residual_5, additional_residual_6, additional_residual_7, additional_residual_8, additional_residual_9); +} \ No newline at end of file