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
950e1bb
·
verified ·
1 Parent(s): e4eaa96

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-189000/config.json +79 -0
  3. checkpoint-189000/generation_config.json +6 -0
  4. checkpoint-189000/model.safetensors +3 -0
  5. checkpoint-189000/optimizer.pt +3 -0
  6. checkpoint-189000/rng_state_0.pth +3 -0
  7. checkpoint-189000/rng_state_1.pth +3 -0
  8. checkpoint-189000/rng_state_2.pth +3 -0
  9. checkpoint-189000/rng_state_3.pth +3 -0
  10. checkpoint-189000/rng_state_4.pth +3 -0
  11. checkpoint-189000/rng_state_5.pth +3 -0
  12. checkpoint-189000/rng_state_6.pth +3 -0
  13. checkpoint-189000/rng_state_7.pth +3 -0
  14. checkpoint-189000/scheduler.pt +3 -0
  15. checkpoint-189000/trainer_state.json +3 -0
  16. checkpoint-189000/training_args.bin +3 -0
  17. checkpoint-190000/config.json +79 -0
  18. checkpoint-190000/generation_config.json +6 -0
  19. checkpoint-190000/model.safetensors +3 -0
  20. checkpoint-190000/optimizer.pt +3 -0
  21. checkpoint-190000/rng_state_0.pth +3 -0
  22. checkpoint-190000/rng_state_1.pth +3 -0
  23. checkpoint-190000/rng_state_2.pth +3 -0
  24. checkpoint-190000/rng_state_3.pth +3 -0
  25. checkpoint-190000/rng_state_4.pth +3 -0
  26. checkpoint-190000/rng_state_5.pth +3 -0
  27. checkpoint-190000/rng_state_6.pth +3 -0
  28. checkpoint-190000/rng_state_7.pth +3 -0
  29. checkpoint-190000/scheduler.pt +3 -0
  30. checkpoint-190000/trainer_state.json +3 -0
  31. checkpoint-190000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -64,3 +64,5 @@ checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  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
 
 
 
64
  checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  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
checkpoint-189000/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-189000/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-189000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56fa6b0650711ef375f17ed116b57bb8a4ba213390145c813ae036b01b542970
3
+ size 4411759432
checkpoint-189000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8865d9a4ecd5c750b72d5f2e1655bc7cc3b22e17e0b74ad29e341e8cab4552c
3
+ size 6332050591
checkpoint-189000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3da430590adf952ec5d20b0a7ddaf6f3fd9abdecd95db5f7e6317d96e13c57c9
3
+ size 16389
checkpoint-189000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb9ab57db999857bed51cf7bd1cba4db2d713cefdfbdb1cb57a6e3fe9b9874de
3
+ size 16389
checkpoint-189000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79be485dc53daed02d83a480cfcc9587a31b4089e484fd081e80865d2b53c9d1
3
+ size 16389
checkpoint-189000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:189fcf080f86696a2e281699dd7da6ff7c072e033491aaa129b304ba66c2237d
3
+ size 16389
checkpoint-189000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d421a5dcd8783e53cf4b61e35e122d35eb0dc7290a1dffd3205fa9fe15eb731
3
+ size 16389
checkpoint-189000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:861001b55ee4a0eed7431b69fca344aa13aacdc0a5d772a1b66ba4f348adfc50
3
+ size 16389
checkpoint-189000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:912e578f854ae97a896c74cedc4c358871a2349ce17c0d88510b66f778795d7d
3
+ size 16389
checkpoint-189000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da60825fc3db983a42f2bcf639d56d8efcaf3a2b5a46e28731064e2080ffb5c9
3
+ size 16389
checkpoint-189000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f970f2d7438f4e58655fce1cddf2f041a6db5b3aa5252bb298d1a8f766dd4771
3
+ size 1465
checkpoint-189000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbf14496d5cc505565d4d03be36582c640ebca6e9103932a2e4130d7b4d00ed8
3
+ size 29067178
checkpoint-189000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ad73daf1da46825f7f55465abc6534b87335f6c1a5756cd2bd808aeca86e859
3
+ size 6481
checkpoint-190000/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-190000/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-190000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bad41d51d1adef39a0169ee09a0a51139c2fb0a562021c6c023ce1df020bbda8
3
+ size 4411759432
checkpoint-190000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36aae5f03104ed782cf501c1ad5e1286f25bf173a66265ee469ba53e792aa84d
3
+ size 6332050591
checkpoint-190000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0e8e6fde7dae1a270280128f28970299d93f1caafe8c413a3664f3dd59ee00f
3
+ size 16389
checkpoint-190000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9e55c862fcd178f4af38127f8db14d0388e2c3ce5054a840727920ebf23ed4b
3
+ size 16389
checkpoint-190000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f7058b46a6c2892e875269cefc52c876e86fdaaffbf4af53cbeb025ed7a4a26
3
+ size 16389
checkpoint-190000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c436d39a60a68d94aeab94cefd2fc80f1f2afa0ff99cb4fa3c639cb7eb01239f
3
+ size 16389
checkpoint-190000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5845f40e5536a2f303e5722e18e234c80d0f9d53ac3b1f8e3cdceb7a5f394c77
3
+ size 16389
checkpoint-190000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85c209807372c182d424f9160a90522d81b15d397bcac826de319d791cba8046
3
+ size 16389
checkpoint-190000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dff759775b021dfd49da3f565db9a434ebac97a876df60951e0f9c52e0991068
3
+ size 16389
checkpoint-190000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:366dc54ca42bbb1380be5ab0cba264ff0a322e2a4d7f5bbb39bf56acb44500df
3
+ size 16389
checkpoint-190000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61661ebcd549b2fb4aadfed629d71722a171e5647d78f0377f0288efd92527c5
3
+ size 1465
checkpoint-190000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41f3ed29b4bb9478f2e6592f4108606a0a279547c7edfec3cf7fb151cac3e5fa
3
+ size 29221571
checkpoint-190000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ad73daf1da46825f7f55465abc6534b87335f6c1a5756cd2bd808aeca86e859
3
+ size 6481