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
8650775
·
verified ·
1 Parent(s): a5b379b

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-149000/config.json +79 -0
  3. checkpoint-149000/generation_config.json +6 -0
  4. checkpoint-149000/model.safetensors +3 -0
  5. checkpoint-149000/optimizer.pt +3 -0
  6. checkpoint-149000/rng_state_0.pth +3 -0
  7. checkpoint-149000/rng_state_1.pth +3 -0
  8. checkpoint-149000/rng_state_2.pth +3 -0
  9. checkpoint-149000/rng_state_3.pth +3 -0
  10. checkpoint-149000/rng_state_4.pth +3 -0
  11. checkpoint-149000/rng_state_5.pth +3 -0
  12. checkpoint-149000/rng_state_6.pth +3 -0
  13. checkpoint-149000/rng_state_7.pth +3 -0
  14. checkpoint-149000/scheduler.pt +3 -0
  15. checkpoint-149000/trainer_state.json +3 -0
  16. checkpoint-149000/training_args.bin +3 -0
  17. checkpoint-150000/config.json +79 -0
  18. checkpoint-150000/generation_config.json +6 -0
  19. checkpoint-150000/model.safetensors +3 -0
  20. checkpoint-150000/optimizer.pt +3 -0
  21. checkpoint-150000/rng_state_0.pth +3 -0
  22. checkpoint-150000/rng_state_1.pth +3 -0
  23. checkpoint-150000/rng_state_2.pth +3 -0
  24. checkpoint-150000/rng_state_3.pth +3 -0
  25. checkpoint-150000/rng_state_4.pth +3 -0
  26. checkpoint-150000/rng_state_5.pth +3 -0
  27. checkpoint-150000/rng_state_6.pth +3 -0
  28. checkpoint-150000/rng_state_7.pth +3 -0
  29. checkpoint-150000/scheduler.pt +3 -0
  30. checkpoint-150000/trainer_state.json +3 -0
  31. checkpoint-150000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -56,3 +56,5 @@ checkpoint-131000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
  checkpoint-132000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-139000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
56
  checkpoint-132000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-139000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
59
+ checkpoint-149000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
60
+ checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-149000/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-149000/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-149000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d5a82f289477a82b6a98b00b0611be8071e4234e903fd06202a8547627b99eb
3
+ size 4411759432
checkpoint-149000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46ddd9876c21dd47ec878cbb9048beb46af20d66d8db580625fa47f2afa9b20e
3
+ size 6332050591
checkpoint-149000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf0080ab396802bced98dbacbc10834c786f92abfee7972f8ed7711bfb833ec8
3
+ size 16389
checkpoint-149000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e56f065ed60ebcf06a6b168f89ed8e603b961e6fb03b632ddd683bdc0ceff11e
3
+ size 16389
checkpoint-149000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39f82a5adf466643cbfdb26dd3ad9b012c81528d417cccccdcf88e09db6f93a6
3
+ size 16389
checkpoint-149000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2958eb0f5ccadc118fbca5f9c34afd919612fd51c4772b0509c3205266377739
3
+ size 16389
checkpoint-149000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:675f8479fe53396ef5fe9ba1f1fd60bf918223874abe12d7666e864d489faeca
3
+ size 16389
checkpoint-149000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a74db4c029d16092156b035458cc1c1f18bff5cabdafb11c8d8add8bcbeeee03
3
+ size 16389
checkpoint-149000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:582e1e168b3df6fbd7f59f521be165fe6f1c6eed0890d95c26f1540f1f72d419
3
+ size 16389
checkpoint-149000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a41445713674769ddab4c195e8ec80a1275c30b05067969aa99b0e1415ccaf7b
3
+ size 16389
checkpoint-149000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71e4fe8ec9680480c8e39d699583bcea88d01a4f2a18819759e6d06121ec1c08
3
+ size 1465
checkpoint-149000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ef467140d98832ff3d42c37f888cb74df7e14910d41353d79bc8dbb71adde7e
3
+ size 22889931
checkpoint-149000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7eaef124a970122a37b1f28e458194b938b928ba59786d0d9adfa0c3cf178fd0
3
+ size 6481
checkpoint-150000/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-150000/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-150000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef577de268154605f85a6631013f37e9018c0cf92b2cea93c969886ef3ba7026
3
+ size 4411759432
checkpoint-150000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:924d9983818f667b414f03a5f8f8baee7cfd145d5c06f3b60165f33c34681fa5
3
+ size 6332050591
checkpoint-150000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f11b6197d1ca8734929cdb5b5cfe6baec5be21bce72d8c2517347e19c9352375
3
+ size 16389
checkpoint-150000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbba4ae0db69c8be2558e8628143e4079492eeaaf0f9ea2b8dae3686191602d0
3
+ size 16389
checkpoint-150000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d82b757c41d5e1252053b4723b1239c3e6b87ea7af9b3a6bcfe9ff392376153
3
+ size 16389
checkpoint-150000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ab62ff98f8949458dcab16dc18c867c6a0088f215a987b08151c91a2ae19cf2
3
+ size 16389
checkpoint-150000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b45c0beaee253f8b4c7ef2e6ce968a4b7e5d85a30e9f1eb1c170200e7854ae0
3
+ size 16389
checkpoint-150000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b0d62abb9fe6b830622283565c4021265b715cca38c396198901ce92a7e44b9
3
+ size 16389
checkpoint-150000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:035593d54e6674a276f44b8222e849f7e8a0f99921af58f148aba5ddd46b9ad2
3
+ size 16389
checkpoint-150000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d691cb8cc3fa2ca35f64c2566ee18ac94fa61e6c8fac6002e1caab450e77a06
3
+ size 16389
checkpoint-150000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bafe711a1976fdd22d9f4ff9762ac139c2dbf1caf9c0ed74bbf52f8154374ca8
3
+ size 1465
checkpoint-150000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0f569dc431634cf29dcea8b6b30b8bfa19ebf24b127a3edef399cfc42d0af4e
3
+ size 23044627
checkpoint-150000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7eaef124a970122a37b1f28e458194b938b928ba59786d0d9adfa0c3cf178fd0
3
+ size 6481