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
313958b
·
verified ·
1 Parent(s): 8d8898a

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-169000/config.json +79 -0
  3. checkpoint-169000/generation_config.json +6 -0
  4. checkpoint-169000/model.safetensors +3 -0
  5. checkpoint-169000/optimizer.pt +3 -0
  6. checkpoint-169000/rng_state_0.pth +3 -0
  7. checkpoint-169000/rng_state_1.pth +3 -0
  8. checkpoint-169000/rng_state_2.pth +3 -0
  9. checkpoint-169000/rng_state_3.pth +3 -0
  10. checkpoint-169000/rng_state_4.pth +3 -0
  11. checkpoint-169000/rng_state_5.pth +3 -0
  12. checkpoint-169000/rng_state_6.pth +3 -0
  13. checkpoint-169000/rng_state_7.pth +3 -0
  14. checkpoint-169000/scheduler.pt +3 -0
  15. checkpoint-169000/trainer_state.json +3 -0
  16. checkpoint-169000/training_args.bin +3 -0
  17. checkpoint-170000/config.json +79 -0
  18. checkpoint-170000/generation_config.json +6 -0
  19. checkpoint-170000/model.safetensors +3 -0
  20. checkpoint-170000/optimizer.pt +3 -0
  21. checkpoint-170000/rng_state_0.pth +3 -0
  22. checkpoint-170000/rng_state_1.pth +3 -0
  23. checkpoint-170000/rng_state_2.pth +3 -0
  24. checkpoint-170000/rng_state_3.pth +3 -0
  25. checkpoint-170000/rng_state_4.pth +3 -0
  26. checkpoint-170000/rng_state_5.pth +3 -0
  27. checkpoint-170000/rng_state_6.pth +3 -0
  28. checkpoint-170000/rng_state_7.pth +3 -0
  29. checkpoint-170000/scheduler.pt +3 -0
  30. checkpoint-170000/trainer_state.json +3 -0
  31. checkpoint-170000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -60,3 +60,5 @@ 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
61
  checkpoint-159000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
60
  checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-159000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
+ 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
checkpoint-169000/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-169000/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-169000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8367b075af291a25eca877ffe37282074fee9c4b84168393455341fb3283d0fa
3
+ size 4411759432
checkpoint-169000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a78a8ba4d9a9b4f1c7bcc0b0ece630b57816ad318e0e54a6621b397d4b51b813
3
+ size 6332050591
checkpoint-169000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd3bba64314ddcf2b007e00959610559a902343e6af0d7c42189756439382c1d
3
+ size 16389
checkpoint-169000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5db566cd21be2f6f84557d0ec50b41a58e0ec893a3fb4f9824aeae0c63fedb7
3
+ size 16389
checkpoint-169000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38b1f2686660f77a06da403121a406d6aa7f57ba1b24a419453d1e4039c51bc2
3
+ size 16389
checkpoint-169000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ba46dcd530de926dacb8603e02316cde2a2081e0c497fa0c7aa132af9a61780
3
+ size 16389
checkpoint-169000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8afcad8217c9d6d1c8076cdf1a5a9848282d3230940c1a8b831e11de8939af69
3
+ size 16389
checkpoint-169000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a111b22f84093419d77b975d92d19a7eab21d2294af0841bfa200445ea4ce66
3
+ size 16389
checkpoint-169000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:525b43e873344f6740a689534b16970256436205fd56308399a61964a0fdc803
3
+ size 16389
checkpoint-169000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee0685c9b3e9318fa67b67383aa1376cbf1229cdb002f9b0504e7949d9a5c908
3
+ size 16389
checkpoint-169000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7302bef595c2dd408e8c1eae13a8d09eab77e9480e94c0c9508009ffe3596533
3
+ size 1465
checkpoint-169000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bba7434f6f85ba5471a0387ab2b83b0112caa2d1b7d15d698716427b999e729
3
+ size 25980361
checkpoint-169000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db00cf522d239896ac36cde524f4d4593da6fcf0534c62c64f4f3aac0aa7bae9
3
+ size 6481
checkpoint-170000/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-170000/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-170000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12325e3aed3e9cf3c6c4d04cea0e70ecff2a5aa340e84c3074b20bf58e750acb
3
+ size 4411759432
checkpoint-170000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d34476804ffec222b01cbd834cbd99d010c391e0b70591c682b24341d33eaf65
3
+ size 6332050591
checkpoint-170000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67520256f37ff2732fecf22c96c539af3b9af8677d6baeab29f77c181baad655
3
+ size 16389
checkpoint-170000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3730c1297882064b230f0f2006195696b95e24e98a02453afc96c394faf6c4c2
3
+ size 16389
checkpoint-170000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d72a6b793530a3892d2c96a92d2038e7219e5d472ea40648a332d866d8509d29
3
+ size 16389
checkpoint-170000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c8410000b3661539a528975359673047121febb686c9a2839ba0edea46103d5
3
+ size 16389
checkpoint-170000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcfacc9f3610e1a26bf706ab4a9207026257d576087a1f2d105fdc1aa7dc746f
3
+ size 16389
checkpoint-170000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01111d623eda6894afaaf521806e9489b5295aeb38d21e941d8ef85dea514249
3
+ size 16389
checkpoint-170000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f881f7b01e73ce09ed111588ecf64f3ae4856430e235acee24f0cd831d97f265
3
+ size 16389
checkpoint-170000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b00028c0ce82716a4933c3efa57a81e486d53b3dd26cd2f7540868f8db495944
3
+ size 16389
checkpoint-170000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c42b7c0c2872d28738a7dc651cc057faf3183fe07b253ae36d534d0b294ab22
3
+ size 1465
checkpoint-170000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de3a860570264cc615cc3660fa3fd66ca85231ce79eb0dd6d2c6b358b8e2879d
3
+ size 26134715
checkpoint-170000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db00cf522d239896ac36cde524f4d4593da6fcf0534c62c64f4f3aac0aa7bae9
3
+ size 6481