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
7d2ae3a
·
verified ·
1 Parent(s): 7f1c23f

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-89000/config.json +79 -0
  3. checkpoint-89000/generation_config.json +6 -0
  4. checkpoint-89000/model.safetensors +3 -0
  5. checkpoint-89000/optimizer.pt +3 -0
  6. checkpoint-89000/rng_state_0.pth +3 -0
  7. checkpoint-89000/rng_state_1.pth +3 -0
  8. checkpoint-89000/rng_state_2.pth +3 -0
  9. checkpoint-89000/rng_state_3.pth +3 -0
  10. checkpoint-89000/rng_state_4.pth +3 -0
  11. checkpoint-89000/rng_state_5.pth +3 -0
  12. checkpoint-89000/rng_state_6.pth +3 -0
  13. checkpoint-89000/rng_state_7.pth +3 -0
  14. checkpoint-89000/scheduler.pt +3 -0
  15. checkpoint-89000/trainer_state.json +3 -0
  16. checkpoint-89000/training_args.bin +3 -0
  17. checkpoint-90000/config.json +79 -0
  18. checkpoint-90000/generation_config.json +6 -0
  19. checkpoint-90000/model.safetensors +3 -0
  20. checkpoint-90000/optimizer.pt +3 -0
  21. checkpoint-90000/rng_state_0.pth +3 -0
  22. checkpoint-90000/rng_state_1.pth +3 -0
  23. checkpoint-90000/rng_state_2.pth +3 -0
  24. checkpoint-90000/rng_state_3.pth +3 -0
  25. checkpoint-90000/rng_state_4.pth +3 -0
  26. checkpoint-90000/rng_state_5.pth +3 -0
  27. checkpoint-90000/rng_state_6.pth +3 -0
  28. checkpoint-90000/rng_state_7.pth +3 -0
  29. checkpoint-90000/scheduler.pt +3 -0
  30. checkpoint-90000/trainer_state.json +3 -0
  31. checkpoint-90000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -40,3 +40,5 @@ checkpoint-79000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
40
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
  checkpoint-87000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
40
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
  checkpoint-87000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
+ checkpoint-89000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
+ checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-89000/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-89000/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-89000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5830ebb14d72024107dd03f3d9983463bbe391d90723d8d9d0e017dc851c2938
3
+ size 4411759432
checkpoint-89000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:400d89cd18c69a70f16be20910948aaec68e336e5768934931228c56421fb736
3
+ size 6332050591
checkpoint-89000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19938b1075ca0eb086e193ad5dc3a1def6636011a77d4de8b588fb27288cdc76
3
+ size 16389
checkpoint-89000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51ae5c969e69a6eb4966e16510816813e6279b76319890c1ed1596b8186e2648
3
+ size 16389
checkpoint-89000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:738003d19df67aa4ad64e3e9f675c0bb9a7f5d47348120f1e3539183219e2411
3
+ size 16389
checkpoint-89000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f7de32b373892785330c877fe562f48789ae39c9dc2875c42eda1d348580f02
3
+ size 16389
checkpoint-89000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e30aed288ac96655e6867995858827bbddf28896c7620a4ad7010b7593806d2
3
+ size 16389
checkpoint-89000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5625dfb3c3e716755dae2686ff6ecf405dfcb990132911a56c752af47dea179
3
+ size 16389
checkpoint-89000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b8b7cf2fb3aef54311973e4f914713590c5af5c8c3af02677de52942bb12613
3
+ size 16389
checkpoint-89000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18bc0c9cc3a96c91104d932f989f78f5d7f08b42f5e1ebcde1c7d8da5d80c472
3
+ size 16389
checkpoint-89000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:365708b626ab580ff59aee928ce892389c364422d21cf662e71dd1a09455fafc
3
+ size 1465
checkpoint-89000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f726ed590e0995f80e0832e643e7016edfa6ac97728a1ea523c528b7b07dbe3
3
+ size 13622183
checkpoint-89000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ec52a4b28132bf19db51aa951394b8ea78d93435d46fd867f8f5e263b67c97c
3
+ size 6481
checkpoint-90000/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-90000/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-90000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f193d5a3cdc93bf424a864bc95231ebd96ea13ce3f278b622f40bca6bf58e4c4
3
+ size 4411759432
checkpoint-90000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0c81332b8b18ebad72e29a1f8fe1a1bd97270e8caf49c5fdb3796c8fdc921f9
3
+ size 6332050591
checkpoint-90000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93046b9871d885e40a1c32c1b8f7ee218c7060c36ad3c2991507ecb59576fbd8
3
+ size 16389
checkpoint-90000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d874204ae976d1f2b34d7f4248292af18296de8ad5259050eed109b694e46bd3
3
+ size 16389
checkpoint-90000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbbad7332c7b4f0e317a25d2f77a15027a8527067d63d8d5f386753a16c76f58
3
+ size 16389
checkpoint-90000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61d4d2029dba72bf37e4a283a36087dcdbd18834e8aeadbc3581999f048e8851
3
+ size 16389
checkpoint-90000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bfa6c100fb96a210333962985db5222ac3883acb08a591e5ec50b03ba61c060
3
+ size 16389
checkpoint-90000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4411c8b8e715fc5cebe7c8f6241b306da107d146b5d21c683fa060399220f24d
3
+ size 16389
checkpoint-90000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2abc518e5603dda26fb6e556061f1881e83cdc3b1d5e6bce0b6afcf25289dd59
3
+ size 16389
checkpoint-90000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e16f07ce0bb64b35ae22d2769d6060414fc967242768eb1b46855bc4ff40fcc2
3
+ size 16389
checkpoint-90000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c13aba3641876275d96842977cfe719b3bed43aa2825aa5929f6aeb2f0d0b93c
3
+ size 1465
checkpoint-90000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:641646c67539a0d273afb8bc54289841fcdc55874d7bd6d88aa9ac255d5762c8
3
+ size 13775632
checkpoint-90000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ec52a4b28132bf19db51aa951394b8ea78d93435d46fd867f8f5e263b67c97c
3
+ size 6481