Spaces:
Running
on
Zero
Running
on
Zero
File size: 6,616 Bytes
ca066a9 40a1f37 ca066a9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 |
T5_SHUNT_REPOS = {
"clip_g": {
"models": ["vit-bigG-14", 'flan-t5-base'],
"config": {
"adapter_id": "003", "name": "DualShuntAdapter-G",
"t5": {
"model": "google/flan-t5-base",
"hidden_size": 768
},
"clip": {
"model": "openai/clip-vit-large-patch14",
"hidden_size": 1280
},
"hidden_size": 1280, # This is the adapter's output size
"bottleneck": 640, "heads": 20,
"tau_init": 0.1, "max_guidance": 10.0,
"proj_layers": 2, "layer_norm": True, "dropout": 0.1,
"use_dropout": True, "use_proj_stack": True, "assert_input_dims": True,
"routing": {"type": "cross_attention", "enable_causal_mask": False, "bidirectional": True},
"version": "v0.3.2"
},
"repo": "AbstractPhil/t5-flan-base-vit-bigG-14-dual-stream-adapter",
"shunts_available": {
"shunt_type_name": "DualStreamAdapter-G",
"config_file_name": "config.json",
"shunt_list": [
"t5-flan-vit-bigG-14-dual_shunt_caption.safetensors",
"t5-flan-vit-bigG-14-dual_shunt_no_caption_e1.safetensors",
"t5-flan-vit-bigG-14-dual_shunt_no_caption_e2.safetensors",
"t5-flan-vit-bigG-14-dual_shunt_no_caption_e3.safetensors",
"t5-flan-vit-bigG-14-dual_shunt_summarize.safetensors",
"dual_shunt_omega_no_caption_e1_step_10000.safetensors",
"dual_shunt_omega_no_caption_noised_e1_step_1000.safetensors",
"dual_shunt_omega_no_caption_noised_e1_step_4000.safetensors",
"dual_shunt_omega_no_caption_noised_e1_step_10000.safetensors",
],
}
},
"clip_l": {
"models": ["vit-l-14", 'flan-t5-base'],
"config": {
"adapter_id": "002",
"name": "DualShuntAdapter",
"t5": {"model": "google/flan-t5-base", "hidden_size": 768},
"clip": {"model": "openai/clip-vit-large-patch14", "hidden_size": 768},
"hidden_size": 768, # This is the adapter's output size
"bottleneck": 384, "heads": 12,
"tau_init": 0.1, "max_guidance": 10.0,
"proj_layers": 2, "layer_norm": True, "dropout": 0.1,
"use_dropout": True, "use_proj_stack": True, "assert_input_dims": True,
"routing": {"type": "cross_attention", "enable_causal_mask": False, "bidirectional": True},
"version": "v0.3.2"
},
"repo": "AbstractPhil/t5-flan-base-vit-l-14-dual-stream-adapter",
"shunts_available": {
"shunt_type_name": "DualStreamAdapter-L",
"config_file_name": "config.json",
"shunt_list": [
"t5-vit-l-14-dual_shunt_caption.safetensors",
"t5-vit-l-14-dual_shunt_no_caption.safetensors",
"t5-vit-l-14-dual_shunt_summarize.safetensors",
"t5-vit-l-14-dual_shunt_booru_13_000_000.safetensors",
"t5-vit-l-14-dual_shunt_booru_51_200_000.safetensors",
],
},
}
}
# ─── Adapter Configs ─────────────────────────────────────────────
BERT_CONFIGS = {
"mobilebert-base-uncased": {
"repo_name": "google/mobilebert-uncased",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
"subfolder": "",
},
"bert-base-uncased": {
"repo_name": "bert-base-uncased",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
},
"bert-large-uncased": {
"repo_name": "bert-large-uncased",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
},
"bert-base-cased": {
"repo_name": "bert-base-cased",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
}
}
T5_CONFIGS = {
"flan-t5-base": {
"repo_name": "google/flan-t5-base",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
},
"t5-small": {
"repo_name": "google-t5/t5-small",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
},
"t5_small_human_attentive_try2_pass3": {
"repo_name": "AbstractPhil/t5_small_human_attentive_try2_pass3",
"use_huggingface": True, # defaults to simple loading from HuggingFace, if False, will use repo_name and subfolder
# the necessary config is present here for posterity in case it fails to load from HuggingFace.
"subfolder": "",
"tokenizer": "t5-small",
"file_name": "model.safetensors",
"config": {
"config_file_name": "config.json",
"architectures": [
"T5ForConditionalGeneration"
],
"attention_dropout": 0.3,
"classifier_dropout": 0.0,
"d_ff": 2048,
"d_kv": 64,
"d_model": 512,
"decoder_start_token_id": 0,
"dense_act_fn": "relu",
"dropout_rate": 0.0, #0.3, # disable for generation
"eos_token_id": 1,
"feed_forward_proj": "relu",
"initializer_factor": 1.0,
"is_encoder_decoder": True,
"is_gated_act": False,
"layer_norm_epsilon": 1e-06,
"model_type": "t5",
"n_positions": 512,
"num_decoder_layers": 6,
"num_heads": 8,
"num_layers": 6,
"output_past": True,
"pad_token_id": 0,
"relative_attention_max_distance": 128,
"relative_attention_num_buckets": 32,
"task_specific_params": {
"caption": {
"early_stopping": True,
"length_penalty": 1.0,
"max_length": 64,
"num_beams": 4,
"prefix": "caption: "
}
},
"torch_dtype": "float32",
"transformers_version": "4.51.3",
"use_cache": True,
"vocab_size": 32128
}
}
} |