Image-to-Image
univa
Safetensors
univa_qwen2p5vl
LanguageBind commited on
Commit
5cf38a9
·
verified ·
1 Parent(s): 8c4bcdf

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +103 -0
config.json ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "UnivaQwen2p5VLForConditionalGeneration"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": 151643,
7
+ "decay": 0.999,
8
+ "denoise_tower": {
9
+ "_attn_implementation_autoset": true,
10
+ "denoise_projector_type": "mlp2x_gelu",
11
+ "denoiser_config": {
12
+ "_class_name": "FluxTransformer2DModel",
13
+ "_diffusers_version": "0.30.0.dev0",
14
+ "_name_or_path": "../checkpoints/flux-dev/transformer",
15
+ "attention_head_dim": 128,
16
+ "guidance_embeds": true,
17
+ "in_channels": 64,
18
+ "joint_attention_dim": 4096,
19
+ "num_attention_heads": 24,
20
+ "num_layers": 19,
21
+ "num_single_layers": 38,
22
+ "patch_size": 1,
23
+ "pooled_projection_dim": 768
24
+ },
25
+ "denoiser_type": "flux",
26
+ "input_hidden_size": 3584,
27
+ "model_type": "univa_denoise_tower",
28
+ "output_hidden_size": 4096,
29
+ "torch_dtype": "float32",
30
+ "vae_input_hidden_size": 64,
31
+ "vae_projector_type": "mlp2x_gelu"
32
+ },
33
+ "eos_token_id": 151645,
34
+ "hidden_act": "silu",
35
+ "hidden_size": 3584,
36
+ "image_begin_token_id": 151652,
37
+ "image_end_token_id": 151653,
38
+ "image_token_id": 151655,
39
+ "image_token_length": null,
40
+ "initializer_range": 0.02,
41
+ "intermediate_size": 18944,
42
+ "inv_gamma": 1.0,
43
+ "max_position_embeddings": 128000,
44
+ "max_window_layers": 28,
45
+ "min_decay": 0.0,
46
+ "model_type": "univa_qwen2p5vl",
47
+ "num_attention_heads": 28,
48
+ "num_hidden_layers": 28,
49
+ "num_key_value_heads": 4,
50
+ "optimization_step": 40000,
51
+ "power": 0.6666666666666666,
52
+ "rms_norm_eps": 1e-06,
53
+ "rope_scaling": {
54
+ "mrope_section": [
55
+ 16,
56
+ 24,
57
+ 24
58
+ ],
59
+ "rope_type": "default",
60
+ "type": "default"
61
+ },
62
+ "rope_theta": 1000000.0,
63
+ "shortcut_image_embeds": false,
64
+ "shortcut_image_embeds_scale": 0.5,
65
+ "shortcut_projector_type": null,
66
+ "sliding_window": 32768,
67
+ "tie_word_embeddings": false,
68
+ "torch_dtype": "float32",
69
+ "transformers_version": "4.50.0",
70
+ "update_after_step": 0,
71
+ "use_cache": true,
72
+ "use_ema_warmup": false,
73
+ "use_sliding_window": false,
74
+ "video_token_id": 151656,
75
+ "vision_config": {
76
+ "depth": 32,
77
+ "fullatt_block_indexes": [
78
+ 7,
79
+ 15,
80
+ 23,
81
+ 31
82
+ ],
83
+ "hidden_act": "silu",
84
+ "hidden_size": 1280,
85
+ "in_channels": 3,
86
+ "in_chans": 3,
87
+ "intermediate_size": 3420,
88
+ "model_type": "qwen2_5_vl",
89
+ "num_heads": 16,
90
+ "out_hidden_size": 3584,
91
+ "patch_size": 14,
92
+ "spatial_merge_size": 2,
93
+ "spatial_patch_size": 14,
94
+ "temporal_patch_size": 2,
95
+ "tokens_per_second": 2,
96
+ "torch_dtype": "float32",
97
+ "window_size": 112
98
+ },
99
+ "vision_end_token_id": 151653,
100
+ "vision_start_token_id": 151652,
101
+ "vision_token_id": 151654,
102
+ "vocab_size": 152064
103
+ }