Hugging Face's logo

umd-vt-nyu
/
JH_multinode_dc-ae-f32c32-in-1.0-diffusers-768_patch-1_group-7_5e4_residual_attn

Safetensors
Model card Files Community
jiuhai commited on
Commit
fbfeee9
·
verified ·
1 Parent(s): fd8f14e

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-269000/config.json +79 -0
  3. checkpoint-269000/generation_config.json +6 -0
  4. checkpoint-269000/model.safetensors +3 -0
  5. checkpoint-269000/optimizer.pt +3 -0
  6. checkpoint-269000/rng_state_0.pth +3 -0
  7. checkpoint-269000/rng_state_1.pth +3 -0
  8. checkpoint-269000/rng_state_2.pth +3 -0
  9. checkpoint-269000/rng_state_3.pth +3 -0
  10. checkpoint-269000/rng_state_4.pth +3 -0
  11. checkpoint-269000/rng_state_5.pth +3 -0
  12. checkpoint-269000/rng_state_6.pth +3 -0
  13. checkpoint-269000/rng_state_7.pth +3 -0
  14. checkpoint-269000/scheduler.pt +3 -0
  15. checkpoint-269000/trainer_state.json +3 -0
  16. checkpoint-269000/training_args.bin +3 -0
  17. checkpoint-270000/config.json +79 -0
  18. checkpoint-270000/generation_config.json +6 -0
  19. checkpoint-270000/model.safetensors +3 -0
  20. checkpoint-270000/optimizer.pt +3 -0
  21. checkpoint-270000/rng_state_0.pth +3 -0
  22. checkpoint-270000/rng_state_1.pth +3 -0
  23. checkpoint-270000/rng_state_2.pth +3 -0
  24. checkpoint-270000/rng_state_3.pth +3 -0
  25. checkpoint-270000/rng_state_4.pth +3 -0
  26. checkpoint-270000/rng_state_5.pth +3 -0
  27. checkpoint-270000/rng_state_6.pth +3 -0
  28. checkpoint-270000/rng_state_7.pth +3 -0
  29. checkpoint-270000/scheduler.pt +3 -0
  30. checkpoint-270000/trainer_state.json +3 -0
  31. checkpoint-270000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -82,3 +82,5 @@ checkpoint-249000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
83
  checkpoint-259000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
84
  checkpoint-260000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
82
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
83
  checkpoint-259000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
84
  checkpoint-260000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
85
+ checkpoint-269000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
86
+ checkpoint-270000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-269000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 1,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-in-1.0-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-269000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-269000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:437be321ef26c7a49b8e8e46c86b10af80ed36108bd0aac6f8e1cd76de81dc21
3
+ size 4411759432
checkpoint-269000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4804f08115e3d56a4ba6c9c14817ccb712c73e57832617acbfb876f15c5b44b
3
+ size 6332050591
checkpoint-269000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c435e86e6c93fa91c2e01c4a2f7aebb2d2b96edf59ac31a8540296ace58e117
3
+ size 16389
checkpoint-269000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eaa42211b7452d575536ffc4b81feb85e1a5d0dfe17043fa0ea3ba780d8fa4dc
3
+ size 16389
checkpoint-269000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a182d796f2da84a4b3562cf04dd351bf067d6b6a51e73e21c929b4860096e764
3
+ size 16389
checkpoint-269000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5540b7bf2c9fec5e39031b7acca1a3b2c15ea75db35505461a7a2b97daaabc1
3
+ size 16389
checkpoint-269000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a20b9a42dab66fa57e25686f819b68c65b40d93b21afab6f4fff1ee1dad099bc
3
+ size 16389
checkpoint-269000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b94d6bd9bdbff645ea6b0448e6f82a2ae3b90e016f6b68f100423807d0784714
3
+ size 16389
checkpoint-269000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54241614c053db8c93a7640cc07fa589413403da223daf6946e73db93cec38c3
3
+ size 16389
checkpoint-269000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2795f53ae275b750f8b329b8b74797a6638ee7fefca6c26597df64ecf9135d6
3
+ size 16389
checkpoint-269000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61799ce8717ceddba0353f7900f54cb4ede8d4ded84ee775ac475a46c8f41a28
3
+ size 1465
checkpoint-269000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad3b3c19186849f455c633b53535c573d0b3a7ab71d21b18fd1a2356393acbe0
3
+ size 41414756
checkpoint-269000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df0a089648d99e283b4f0e4890ef82d74de252cadb331feb70bc93fb5759bad6
3
+ size 6481
checkpoint-270000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 1,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-in-1.0-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-270000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-270000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59a7adcdb7d4ed51e3b315a04c6c9aca4978dd9dd4d9bfa5597badac70593323
3
+ size 4411759432
checkpoint-270000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d65a90d5f0bbefab59aa4e5a021e142751e89f7d66cc67f969d9cd85d1721c2b
3
+ size 6332050591
checkpoint-270000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5af1dddd72f8c02b5d8cee10086de6544f30aec05c26d0c5003285ed1dbc0ff6
3
+ size 16389
checkpoint-270000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:534a14dd7ed5ebfbcf676e20e474613cbb6887650faebe75fe45be0871e85d66
3
+ size 16389
checkpoint-270000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d2ba23a39052ad90e52a9facbd4355bec9df733c6a770f59460098ec7a9d1db
3
+ size 16389
checkpoint-270000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afc289a5cf10afd5946bdc811a6f0e7cfce1c206eb08237adfc7d48e4abc99fa
3
+ size 16389
checkpoint-270000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:929d56d8e6af72fac84617f8d952e239b45adacbe037348c81df11fbddc21972
3
+ size 16389
checkpoint-270000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86ccbe4fce3055ce87f9b4faabe575716947453b79169716f97edc08cdc5ec83
3
+ size 16389
checkpoint-270000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6d40e4a9225701bb7bda1ab068918921eb60344c8e44524a3acc48d022f3557
3
+ size 16389
checkpoint-270000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd675221a0ca4b9be94263225d31e224c7b3a7621a1f988e1d2f357e17070e13
3
+ size 16389
checkpoint-270000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd19092349f65db3c42cf17da0c410db58a805531aabca006adc94370faa3e63
3
+ size 1465
checkpoint-270000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66ee6a0c35595fe58176731f05390d7fc024088195e7934f8c100bf199e28b6d
3
+ size 41569132
checkpoint-270000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df0a089648d99e283b4f0e4890ef82d74de252cadb331feb70bc93fb5759bad6
3
+ size 6481