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
97172a0
·
verified ·
1 Parent(s): 950e1bb

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-193000/config.json +79 -0
  3. checkpoint-193000/generation_config.json +6 -0
  4. checkpoint-193000/model.safetensors +3 -0
  5. checkpoint-193000/optimizer.pt +3 -0
  6. checkpoint-193000/rng_state_0.pth +3 -0
  7. checkpoint-193000/rng_state_1.pth +3 -0
  8. checkpoint-193000/rng_state_2.pth +3 -0
  9. checkpoint-193000/rng_state_3.pth +3 -0
  10. checkpoint-193000/rng_state_4.pth +3 -0
  11. checkpoint-193000/rng_state_5.pth +3 -0
  12. checkpoint-193000/rng_state_6.pth +3 -0
  13. checkpoint-193000/rng_state_7.pth +3 -0
  14. checkpoint-193000/scheduler.pt +3 -0
  15. checkpoint-193000/trainer_state.json +3 -0
  16. checkpoint-193000/training_args.bin +3 -0
  17. checkpoint-194000/config.json +79 -0
  18. checkpoint-194000/generation_config.json +6 -0
  19. checkpoint-194000/model.safetensors +3 -0
  20. checkpoint-194000/optimizer.pt +3 -0
  21. checkpoint-194000/rng_state_0.pth +3 -0
  22. checkpoint-194000/rng_state_1.pth +3 -0
  23. checkpoint-194000/rng_state_2.pth +3 -0
  24. checkpoint-194000/rng_state_3.pth +3 -0
  25. checkpoint-194000/rng_state_4.pth +3 -0
  26. checkpoint-194000/rng_state_5.pth +3 -0
  27. checkpoint-194000/rng_state_6.pth +3 -0
  28. checkpoint-194000/rng_state_7.pth +3 -0
  29. checkpoint-194000/scheduler.pt +3 -0
  30. checkpoint-194000/trainer_state.json +3 -0
  31. checkpoint-194000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -66,3 +66,5 @@ checkpoint-179000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
66
  checkpoint-180000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
67
  checkpoint-189000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
68
  checkpoint-190000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
66
  checkpoint-180000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
67
  checkpoint-189000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
68
  checkpoint-190000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
69
+ checkpoint-193000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
70
+ checkpoint-194000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-193000/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-193000/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-193000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0dc587b99ef050b62923e3b66a744e4825d0cb93b684fded4b6b8d4e6d4fc09b
3
+ size 4411759432
checkpoint-193000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fbf4bbf5311414682fb7e305aaef27c98d041154e1a1302a12636cfb52ae18a0
3
+ size 6332050591
checkpoint-193000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5cf16aed9e253b76c4832eb5d7656f5e82d1db32707b2744bbbc53500e426eb1
3
+ size 16389
checkpoint-193000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99bdb57f0cd8fd286a5cf02f1c0cca5446c866d814ddcd0dbbe01eb44cf3fab2
3
+ size 16389
checkpoint-193000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f217b72d73863fa511a8ebfef754695d09bb7a4d1ecc56b879ae3f5e94c7a194
3
+ size 16389
checkpoint-193000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ccf2cf046e64bd293c8eb373d9d813083692f818bfafac8a21741e29e233e2a3
3
+ size 16389
checkpoint-193000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e28e35390bb7f3c361ab1b670255c389fb86358f94010907d344e249bd65679
3
+ size 16389
checkpoint-193000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40cc9969880476371f4fba4a56ccde389a48130324c1a63e5d859560b922f56
3
+ size 16389
checkpoint-193000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c459b25a0d73a997e4ca97d22c9a448c63847ed6139acb688e56c56dbc00f4d
3
+ size 16389
checkpoint-193000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89fa5f02d21a8dffe8f29430d3f2cb78048d87790677fea7c0cf55d2c1d58c9f
3
+ size 16389
checkpoint-193000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48fab84892951eb8eb5e49367b69138d982914e15d7782a1ec8e411f465b4ace
3
+ size 1465
checkpoint-193000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8074958fd99d7b58e53428147a01bd22e7f7873091a95311f39b492454d18f34
3
+ size 29684584
checkpoint-193000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8427f8afdafdd76696a7df96d9a71d1490cd281ff5abc881002ef87e82dfe767
3
+ size 6481
checkpoint-194000/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-194000/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-194000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0d8cdcd75aa5e869f808761f30ee518f60eea12e51d9dc2864c96e2679f1563
3
+ size 4411759432
checkpoint-194000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68d4d0b64a8e6e7ebe5f026535bf5beff1265206516d149ea18cdf0bba19d5bf
3
+ size 6332050591
checkpoint-194000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebe1615cd65638094422face93c03a3ca53d8409b28a084fe18a77aeb617a8f6
3
+ size 16389
checkpoint-194000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:099cee5f97c000821a943d96a2fe4bd6e0d617b66fa1410f8341bf3ad0c6f8ca
3
+ size 16389
checkpoint-194000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8aa875f89fe63a00718a29672aeb70c53ca1fd82e7e55888be344748cbe6874e
3
+ size 16389
checkpoint-194000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e014c3d2d417ca07b2060cbf992c0bff968bc5ef4dba72090fbe48e63ebdfaf
3
+ size 16389
checkpoint-194000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e26191ecb2c574e2d3dac02b7bbd4d297286bfb0adfcc7886abf2e4beacd45c
3
+ size 16389
checkpoint-194000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:849b896d1ffc92decc505e9f702ea3f01c08e74d29f8ca4dd3c40776643f68ce
3
+ size 16389
checkpoint-194000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b8b88767588adbe088dbbb504182a83f85cfadb16fbdef320338c942e4dd1a2
3
+ size 16389
checkpoint-194000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a5fb70fefcf524f5c3d4ebb07700b27a68a4bbd93394a419380e2886dbbdbe5
3
+ size 16389
checkpoint-194000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1065b2409bbe91a8df370600623dbabb6ee1021cdf5bca2987a53d7e2d754e40
3
+ size 1465
checkpoint-194000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d418a017f1776b9861eac8f89651e34b911f7c4a40593b2ec1fee3e67c6f57f
3
+ size 29838876
checkpoint-194000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1834fccdd28852465658b7e95d76e664d5399f4212d15be0a589aaa761ac4ad5
3
+ size 6417