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
a5b379b
·
verified ·
1 Parent(s): 2a95247

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-139000/config.json +79 -0
  3. checkpoint-139000/generation_config.json +6 -0
  4. checkpoint-139000/model.safetensors +3 -0
  5. checkpoint-139000/optimizer.pt +3 -0
  6. checkpoint-139000/rng_state_0.pth +3 -0
  7. checkpoint-139000/rng_state_1.pth +3 -0
  8. checkpoint-139000/rng_state_2.pth +3 -0
  9. checkpoint-139000/rng_state_3.pth +3 -0
  10. checkpoint-139000/rng_state_4.pth +3 -0
  11. checkpoint-139000/rng_state_5.pth +3 -0
  12. checkpoint-139000/rng_state_6.pth +3 -0
  13. checkpoint-139000/rng_state_7.pth +3 -0
  14. checkpoint-139000/scheduler.pt +3 -0
  15. checkpoint-139000/trainer_state.json +3 -0
  16. checkpoint-139000/training_args.bin +3 -0
  17. checkpoint-140000/config.json +79 -0
  18. checkpoint-140000/generation_config.json +6 -0
  19. checkpoint-140000/model.safetensors +3 -0
  20. checkpoint-140000/optimizer.pt +3 -0
  21. checkpoint-140000/rng_state_0.pth +3 -0
  22. checkpoint-140000/rng_state_1.pth +3 -0
  23. checkpoint-140000/rng_state_2.pth +3 -0
  24. checkpoint-140000/rng_state_3.pth +3 -0
  25. checkpoint-140000/rng_state_4.pth +3 -0
  26. checkpoint-140000/rng_state_5.pth +3 -0
  27. checkpoint-140000/rng_state_6.pth +3 -0
  28. checkpoint-140000/rng_state_7.pth +3 -0
  29. checkpoint-140000/scheduler.pt +3 -0
  30. checkpoint-140000/trainer_state.json +3 -0
  31. checkpoint-140000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -54,3 +54,5 @@ checkpoint-129000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  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
 
 
 
54
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  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
checkpoint-139000/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-139000/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-139000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f578837d843053faa41e992b1d99304b141979523a96d8ac142092341a1732a
3
+ size 4411759432
checkpoint-139000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54445adb6565ab720d5bdbfdf46073747a144446002740f15de1f6187970fe1b
3
+ size 6332050591
checkpoint-139000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6392155e979d4a390777b75c6f3fe3a2b84f167edd6c9614fba50718403a8a34
3
+ size 16389
checkpoint-139000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e61ff4a91a0333b86c2079245b422efd6e8b4237a237d88a270755972b7d251
3
+ size 16389
checkpoint-139000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddec14660273412d0ec740230e60dd0949fbd59ed1b1a5dc33b771362e3e5bb5
3
+ size 16389
checkpoint-139000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46d965a8e3166903b3e85e8fab59019968660d6bd1262d676db325da8bab0d2b
3
+ size 16389
checkpoint-139000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d8b51504f60807cc8d0178c88da924c4e055c610e2109a2abb648cec85d657f
3
+ size 16389
checkpoint-139000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c1579f8a2db5959f1158ab50019b12d13d4f098ebb41d190b6cf2d35acc9bd6
3
+ size 16389
checkpoint-139000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f6b67101ecbd1136e1433885bb155179fb30627bed1f0b64e468c30c868b250
3
+ size 16389
checkpoint-139000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de82fac17eae8e0b31f626e3ed9a7b02e6935aca6af669d1e6329991014d2420
3
+ size 16389
checkpoint-139000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f7084edd1cfd3dd81f9efc655aaeefbdb01487b1702c21e5c80d104a19f49e9
3
+ size 1465
checkpoint-139000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da216399a27c049c594b1c0bdb755b7d07bb16c35839c11946403d2006b76c4d
3
+ size 21343334
checkpoint-139000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78cbd7db303f23e6564bde50d3e8a5260ba89b20aa6bcc28451b1cd302fef5fa
3
+ size 6481
checkpoint-140000/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-140000/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-140000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8880f3758334f9471f2169cce53df770804d2be479026abfef75c27e4d8aa7ff
3
+ size 4411759432
checkpoint-140000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44cc0ff38ae53b7780c8e21c6b913ed165e155b6ba3c1e5e0e1acd90ed992877
3
+ size 6332050591
checkpoint-140000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27a1b29565687678fb6c48899fe967d3fdcfb0326ff50e4af2a6d39668e0dafb
3
+ size 16389
checkpoint-140000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1617ea6ce38478e7e3b48e1e35289e6761df1a211b3e5441aef86816f0c7015
3
+ size 16389
checkpoint-140000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36cfe9d512a7b9f357a22fd70828a29705598edc1123fd2c5dc758964677b81f
3
+ size 16389
checkpoint-140000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c39cf143a209db864d2146f4e6379cff613b31b68baf62656dca55d6648085a
3
+ size 16389
checkpoint-140000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55c71f58b42d6baa1e0dfdd2fae7ca388278c533028e8828a5a0c9799f1dfced
3
+ size 16389
checkpoint-140000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f79d72c6068d744ce11001d66974fc64a470e605feb8469bf76e1aef6081fc4b
3
+ size 16389
checkpoint-140000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89b20b1545e682c0b97c415136e2012ca99bba86b5195eb334320d715836133a
3
+ size 16389
checkpoint-140000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ff04ce6f3b2282ad93dd776b29e6765aa35ac7abe08f8273db69c5c1b42ac24
3
+ size 16389
checkpoint-140000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6095d814318301060256c558c0db9b1356c0ffbca53208963da8d4ab12bf085d
3
+ size 1465
checkpoint-140000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:679107b26e88a9ced5edb83b35e68837f7036059921174c01b4d02f7fe163453
3
+ size 21498084
checkpoint-140000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78cbd7db303f23e6564bde50d3e8a5260ba89b20aa6bcc28451b1cd302fef5fa
3
+ size 6481