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
xet
Community
jiuhai commited on
Commit
9599d08
·
verified ·
1 Parent(s): 2c9183b

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-289000/config.json +79 -0
  3. checkpoint-289000/generation_config.json +6 -0
  4. checkpoint-289000/model.safetensors +3 -0
  5. checkpoint-289000/optimizer.pt +3 -0
  6. checkpoint-289000/rng_state_0.pth +3 -0
  7. checkpoint-289000/rng_state_1.pth +3 -0
  8. checkpoint-289000/rng_state_2.pth +3 -0
  9. checkpoint-289000/rng_state_3.pth +3 -0
  10. checkpoint-289000/rng_state_4.pth +3 -0
  11. checkpoint-289000/rng_state_5.pth +3 -0
  12. checkpoint-289000/rng_state_6.pth +3 -0
  13. checkpoint-289000/rng_state_7.pth +3 -0
  14. checkpoint-289000/scheduler.pt +3 -0
  15. checkpoint-289000/trainer_state.json +3 -0
  16. checkpoint-289000/training_args.bin +3 -0
  17. checkpoint-290000/config.json +79 -0
  18. checkpoint-290000/generation_config.json +6 -0
  19. checkpoint-290000/model.safetensors +3 -0
  20. checkpoint-290000/optimizer.pt +3 -0
  21. checkpoint-290000/rng_state_0.pth +3 -0
  22. checkpoint-290000/rng_state_1.pth +3 -0
  23. checkpoint-290000/rng_state_2.pth +3 -0
  24. checkpoint-290000/rng_state_3.pth +3 -0
  25. checkpoint-290000/rng_state_4.pth +3 -0
  26. checkpoint-290000/rng_state_5.pth +3 -0
  27. checkpoint-290000/rng_state_6.pth +3 -0
  28. checkpoint-290000/rng_state_7.pth +3 -0
  29. checkpoint-290000/scheduler.pt +3 -0
  30. checkpoint-290000/trainer_state.json +3 -0
  31. checkpoint-290000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -86,3 +86,5 @@ checkpoint-269000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
86
  checkpoint-270000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
87
  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
 
 
 
86
  checkpoint-270000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
87
  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
checkpoint-289000/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-289000/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-289000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8e9d8620b3c190b66679a82d6d8f96339e41506d53a19e52dad0daff5a38906
3
+ size 4411759432
checkpoint-289000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a131e1a31a10eb63a337ba3740d945e477b4019cf024cba78f17d724a88f9f4
3
+ size 6332050591
checkpoint-289000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da676055e261b3bbb9d5fbd7bac7cab1ad5a7778cf0a2b57938a678d7e0deb93
3
+ size 16389
checkpoint-289000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbc512c8219b52966a885c26236aba58e565c9f99d8483bc23868b2fc62f0ed1
3
+ size 16389
checkpoint-289000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d014b619874cad18da932da0201552ebdafca307731f3fe4cfb760ac177a7ba2
3
+ size 16389
checkpoint-289000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:476fd271495979ea162853d81506bfc2d471d0fca04253442f0b283cd1ea98b3
3
+ size 16389
checkpoint-289000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3351bc25576cdd61e0d503f504eab604808c169a58dd0e61f945a9316f94a3bc
3
+ size 16389
checkpoint-289000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28cb1ed8280b88ae668cbe3e1a3ea8bba49a8250cb05c4ce4b181925d13510bb
3
+ size 16389
checkpoint-289000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10ff39f287d851fa89d91175046fcb04c6539cc383ed3a7fcaf694c22c48db1b
3
+ size 16389
checkpoint-289000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f5c6902f802cc8fd17374f42149037ed15d669baa5ee38580277c6dabb31b98
3
+ size 16389
checkpoint-289000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b79821e1b231a29940ae9caca18892d8d0fa6776e14ac9ebf56ebe755d878c41
3
+ size 1465
checkpoint-289000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c0d787f412f0c82a9e085fd156d8fd268f97dae94fe6cb753e2ea026401ddf4
3
+ size 44501702
checkpoint-289000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24ca8c57d73fd392194ec2cd1ec07d3bf0e255f717910a550f7db425a7db2377
3
+ size 6481
checkpoint-290000/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-290000/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-290000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:459e7c0889e7a71f079eac3695b5d8a1fda5bf8657904d4d88fd87499103ab20
3
+ size 4411759432
checkpoint-290000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08ffd877210d650159343e8d61078c9a7a2437acf26f833199acf5c4b86d8a2f
3
+ size 6332050591
checkpoint-290000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f4693f0a46213182f14d80b88c9647caf3bf107d1ded1caa5242107524c6962
3
+ size 16389
checkpoint-290000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57613872409b40393a919c128e4ed0d5547ee03fed8f8630951258ecdac30658
3
+ size 16389
checkpoint-290000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70b0fba5cc0e79ab48e54c6110f4204d1796c563d70152127e0930d468e80d8d
3
+ size 16389
checkpoint-290000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a92a7da4bb536f773e4cc9a3df4421c1fe6cf9b3c0d95dd632ed44bfa4d3fb5e
3
+ size 16389
checkpoint-290000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f04dad7fbf5c6939548f9e1705bb8047d051cea9bf3db3114f852b51fea3d01a
3
+ size 16389
checkpoint-290000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6d5cb5ffd614e27fad2a1852462cd1a8504f865c4422cbf909d93321f7dc32e
3
+ size 16389
checkpoint-290000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90504388bd463a74b6e0a3ff934a116057da0708259275bd4b54c8ebd0de8870
3
+ size 16389
checkpoint-290000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93125c07dd35509585ecf5aa78f143b32559f2eef8492418c078f3d94c757aa4
3
+ size 16389
checkpoint-290000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:206053e0d017c83f7ffc51b669b67ca9e0f42aad81d3917b0fa35279dff1c238
3
+ size 1465
checkpoint-290000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c659f0e7e6edcd2127b284dd42ba26b2be2571f12ec4d93c34d59f5d177c9fd
3
+ size 44656046
checkpoint-290000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24ca8c57d73fd392194ec2cd1ec07d3bf0e255f717910a550f7db425a7db2377
3
+ size 6481