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
e2cc3fd
·
verified ·
1 Parent(s): 12f1af8

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-128000/config.json +79 -0
  3. checkpoint-128000/generation_config.json +6 -0
  4. checkpoint-128000/model.safetensors +3 -0
  5. checkpoint-128000/optimizer.pt +3 -0
  6. checkpoint-128000/rng_state_0.pth +3 -0
  7. checkpoint-128000/rng_state_1.pth +3 -0
  8. checkpoint-128000/rng_state_2.pth +3 -0
  9. checkpoint-128000/rng_state_3.pth +3 -0
  10. checkpoint-128000/rng_state_4.pth +3 -0
  11. checkpoint-128000/rng_state_5.pth +3 -0
  12. checkpoint-128000/rng_state_6.pth +3 -0
  13. checkpoint-128000/rng_state_7.pth +3 -0
  14. checkpoint-128000/scheduler.pt +3 -0
  15. checkpoint-128000/trainer_state.json +3 -0
  16. checkpoint-128000/training_args.bin +3 -0
  17. checkpoint-129000/config.json +79 -0
  18. checkpoint-129000/generation_config.json +6 -0
  19. checkpoint-129000/model.safetensors +3 -0
  20. checkpoint-129000/optimizer.pt +3 -0
  21. checkpoint-129000/rng_state_0.pth +3 -0
  22. checkpoint-129000/rng_state_1.pth +3 -0
  23. checkpoint-129000/rng_state_2.pth +3 -0
  24. checkpoint-129000/rng_state_3.pth +3 -0
  25. checkpoint-129000/rng_state_4.pth +3 -0
  26. checkpoint-129000/rng_state_5.pth +3 -0
  27. checkpoint-129000/rng_state_6.pth +3 -0
  28. checkpoint-129000/rng_state_7.pth +3 -0
  29. checkpoint-129000/scheduler.pt +3 -0
  30. checkpoint-129000/trainer_state.json +3 -0
  31. checkpoint-129000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -49,3 +49,5 @@ checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
49
  checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
50
  checkpoint-119000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
49
  checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
50
  checkpoint-119000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
52
+ checkpoint-128000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
53
+ checkpoint-129000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-128000/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-128000/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-128000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abc6fca4dd16dfda569a5b4cf2c7acb4e262e119d10aad0d3a55fb1652aec459
3
+ size 4411759432
checkpoint-128000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a72903c3f872dda967c072bfeef771c3e2b8c21ade0e087daca12fa1bbe27e8b
3
+ size 6332050591
checkpoint-128000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6eb82ff4026abff94f20deb8f3fecb23732100f8ec026435fa499b8261abf7b
3
+ size 16389
checkpoint-128000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f144dcbedb9432c4d1f94cf15de395790b16e8d3b9259c0b3aa60e620b04dbef
3
+ size 16389
checkpoint-128000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eec938b892a8e07ddb1a08d08f5de597ba93d4d4afb085323255f63126f3066
3
+ size 16389
checkpoint-128000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea12bb25a15ff3f59199c6cb947301fa4291bd392bba871e57de840f96f47c17
3
+ size 16389
checkpoint-128000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24ab302fa227b8072c7293616b6325fabbe76c5e68a094e4f15879890c929f27
3
+ size 16389
checkpoint-128000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3eac749437d67de5d479de33f8515b31c985f3f4616230981984feddf02d4e96
3
+ size 16389
checkpoint-128000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2509cf7ff3e5a70413075b1a5b400331dc412d65b6472212bd90355b8c8a32b
3
+ size 16389
checkpoint-128000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd2ba7bd1d442ef4f5955ba63bea14a715fe798bf728d450d12146b873b91d9a
3
+ size 16389
checkpoint-128000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efd2617a3ed441c7d591c07c3c48918a526593924931b8817377ff7278aebfd9
3
+ size 1465
checkpoint-128000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7c20900303d4495f1d27d0fa79943de8819d72af6b659a86d28800368bd26fb
3
+ size 19642139
checkpoint-128000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2a411215786a4156a3d5ca7e6cacbdfc9c285c570f0181688aeb1256c9c5187
3
+ size 6481
checkpoint-129000/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-129000/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-129000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4189939a2f311b160dc7f1bb3587e7db8ef87d8679649e47d730397f793b09e
3
+ size 4411759432
checkpoint-129000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:832e43cb6d07361cf8eb33f884fb4097d36c01a14c653f34051d097ef7f38938
3
+ size 6332050591
checkpoint-129000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:869710327b8742c3a90bfc69af269025adb60db2463705ff54475d610bf9e7fb
3
+ size 16389
checkpoint-129000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b33b9d7c7cb66653c86a411804e4b3f9014ed3a658bc8fac624ae10e693dee6
3
+ size 16389
checkpoint-129000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f061d04ad0ea3f95f1a07b8d70a2f58680a6f6243a2ec30eff8a3952050f29ba
3
+ size 16389
checkpoint-129000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b512cf4d4b2e533f159a0c5ca7fd4cf9a027534d94f71164e98556361a531c0d
3
+ size 16389
checkpoint-129000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b27c0b7c11412bc1d7f0125318e0af5889bd3468a80e5979cbf26639fe815587
3
+ size 16389
checkpoint-129000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7168a8917e22d9c9013b539eab7bc4b17feb9013fdb70f9cf1b90dc7e4d58f5f
3
+ size 16389
checkpoint-129000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7367c62c92097e52b9e93d7aa5d5477c6d7a230ac907fe7fe2023c1dd3d1c1c9
3
+ size 16389
checkpoint-129000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:882c3c72bc7c48871497b6885368e0ab7e3df807f4f6979e6cb5e4f7d7daa552
3
+ size 16389
checkpoint-129000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a27843b506f8b1004f4541009609b788012ada09cf57b9463942d90944702e58
3
+ size 1465
checkpoint-129000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea45ea7389a8c6249a0f91ec8428452b58a3cdbe652753d124689b8dbe719eb7
3
+ size 19796655
checkpoint-129000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:435a737a88373edf49f46d9153689ba81a9f7eb37f64f2ef4c8d3ba6d0a336ff
3
+ size 6417