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
e9ab08d
·
verified ·
1 Parent(s): 9599d08

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-299000/config.json +79 -0
  3. checkpoint-299000/generation_config.json +6 -0
  4. checkpoint-299000/model.safetensors +3 -0
  5. checkpoint-299000/optimizer.pt +3 -0
  6. checkpoint-299000/rng_state_0.pth +3 -0
  7. checkpoint-299000/rng_state_1.pth +3 -0
  8. checkpoint-299000/rng_state_2.pth +3 -0
  9. checkpoint-299000/rng_state_3.pth +3 -0
  10. checkpoint-299000/rng_state_4.pth +3 -0
  11. checkpoint-299000/rng_state_5.pth +3 -0
  12. checkpoint-299000/rng_state_6.pth +3 -0
  13. checkpoint-299000/rng_state_7.pth +3 -0
  14. checkpoint-299000/scheduler.pt +3 -0
  15. checkpoint-299000/trainer_state.json +3 -0
  16. checkpoint-299000/training_args.bin +3 -0
  17. checkpoint-300000/config.json +79 -0
  18. checkpoint-300000/generation_config.json +6 -0
  19. checkpoint-300000/model.safetensors +3 -0
  20. checkpoint-300000/optimizer.pt +3 -0
  21. checkpoint-300000/rng_state_0.pth +3 -0
  22. checkpoint-300000/rng_state_1.pth +3 -0
  23. checkpoint-300000/rng_state_2.pth +3 -0
  24. checkpoint-300000/rng_state_3.pth +3 -0
  25. checkpoint-300000/rng_state_4.pth +3 -0
  26. checkpoint-300000/rng_state_5.pth +3 -0
  27. checkpoint-300000/rng_state_6.pth +3 -0
  28. checkpoint-300000/rng_state_7.pth +3 -0
  29. checkpoint-300000/scheduler.pt +3 -0
  30. checkpoint-300000/trainer_state.json +3 -0
  31. checkpoint-300000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -88,3 +88,5 @@ checkpoint-279000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
88
  checkpoint-280000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
89
  checkpoint-289000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
90
  checkpoint-290000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
88
  checkpoint-280000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
89
  checkpoint-289000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
90
  checkpoint-290000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
91
+ checkpoint-299000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
92
+ checkpoint-300000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-299000/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-299000/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-299000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:800c70be5b9b9a0ecb52366b7dd79f692fdc6c971aa2f091b1e9c2b425475577
3
+ size 4411759432
checkpoint-299000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61a0156d02c77720a8001af02ee3db754a921f0166da601832d28a08c0690cc4
3
+ size 6332050591
checkpoint-299000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf64b8c81bde5e904450dd4819d74583aa76fd03d99a4a1329b0f896e1476160
3
+ size 16389
checkpoint-299000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb4930a294fc00231d2e8cb993177d578ba11b61e1db803b1cf0b7748b1ad414
3
+ size 16389
checkpoint-299000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da51c2506e2031176a1535137b6af4379acb9e43b9871096e411a1e710435529
3
+ size 16389
checkpoint-299000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2cc057ea8491f7b9acb1141b2b623acdd2fc55e6d1acbd5370dae47eb2953d12
3
+ size 16389
checkpoint-299000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eefd11839f2c9d3ec8df733f26a41f4a89302df35030c111affefe84706e9ae2
3
+ size 16389
checkpoint-299000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88d01e267ae8519707e7e36355212339a3e0c4b28c6d051e1b91fec51939e2c5
3
+ size 16389
checkpoint-299000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88592b8c90074fa696897365c9ca6ec5cc4b25f815f0b06a5446cb4a9e151c38
3
+ size 16389
checkpoint-299000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:962a3660aa064d8c3ae761a18cf730e9b8b7f18f2dec9f223333db12d2c02c78
3
+ size 16389
checkpoint-299000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07b72cb7e0903f68f8dbe21bd3ea0358157efd459928d3445b8a8e6ed5ab2ab0
3
+ size 1465
checkpoint-299000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1dd5d229e0ba2e08a6841a57d48958c003a206aa54c8d9da05a98cb0cb24338
3
+ size 46045399
checkpoint-299000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98babef9df124d98dcb377ab0c47b8b41fc4c4bc433e3cf7ee13d6751d47ca4e
3
+ size 6481
checkpoint-300000/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-300000/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-300000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1608cc211eabf43cdff9ab781b9b1768a51f92a028656d1a78f2db1d14b95c39
3
+ size 4411759432
checkpoint-300000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee1e42f612f2b4b2b4a7e2eddf64a97e74d4fc7bd2b8eae9fa95a905f7697a1f
3
+ size 6332050591
checkpoint-300000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:608212005d11c58aaac13441a44c696924187687ad3970bf64fd2b1cddef611e
3
+ size 16389
checkpoint-300000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:690f2e3c43541935b02a2d311c984a448bce7773ea9e5afe2adb998ba74ff073
3
+ size 16389
checkpoint-300000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c546f782ccccde1edab3684bfd118717ca9405f52b9a1b0b6a543a17c1d9d1c
3
+ size 16389
checkpoint-300000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac303b0ec138ebc5556aea52c7e17c45c5442a3447784112a13cc939c9da862c
3
+ size 16389
checkpoint-300000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40f5cd53265cf6bcec4c1c229aadfe7bde6e956ccfa6e315607a91bbed9927f
3
+ size 16389
checkpoint-300000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5eb961a777a1d9ac3af796fab0373bc175f54c2630db32064589e4f2568266eb
3
+ size 16389
checkpoint-300000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1ad227362b64067f95ff573b2877fec866595c71ea95f5ed885f81374d8243c
3
+ size 16389
checkpoint-300000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e33661ad45422d4f5cf9297a3cf5b9537a2aa1791ca09b305125d113e0397e0
3
+ size 16389
checkpoint-300000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a78e29de92f1b2b0f0d44ab7791fc16d98da62b8fa746a7e87863cf307b2cf1e
3
+ size 1465
checkpoint-300000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4dd14a8b9d4deb5f158fff249666c9ddc835f7dc710330f1f7b01dcbf8c07da8
3
+ size 46199739
checkpoint-300000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98babef9df124d98dcb377ab0c47b8b41fc4c4bc433e3cf7ee13d6751d47ca4e
3
+ size 6481