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
acb304d
·
verified ·
1 Parent(s): 97d4998

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-229000/config.json +79 -0
  3. checkpoint-229000/generation_config.json +6 -0
  4. checkpoint-229000/model.safetensors +3 -0
  5. checkpoint-229000/optimizer.pt +3 -0
  6. checkpoint-229000/rng_state_0.pth +3 -0
  7. checkpoint-229000/rng_state_1.pth +3 -0
  8. checkpoint-229000/rng_state_2.pth +3 -0
  9. checkpoint-229000/rng_state_3.pth +3 -0
  10. checkpoint-229000/rng_state_4.pth +3 -0
  11. checkpoint-229000/rng_state_5.pth +3 -0
  12. checkpoint-229000/rng_state_6.pth +3 -0
  13. checkpoint-229000/rng_state_7.pth +3 -0
  14. checkpoint-229000/scheduler.pt +3 -0
  15. checkpoint-229000/trainer_state.json +3 -0
  16. checkpoint-229000/training_args.bin +3 -0
  17. checkpoint-230000/config.json +79 -0
  18. checkpoint-230000/generation_config.json +6 -0
  19. checkpoint-230000/model.safetensors +3 -0
  20. checkpoint-230000/optimizer.pt +3 -0
  21. checkpoint-230000/rng_state_0.pth +3 -0
  22. checkpoint-230000/rng_state_1.pth +3 -0
  23. checkpoint-230000/rng_state_2.pth +3 -0
  24. checkpoint-230000/rng_state_3.pth +3 -0
  25. checkpoint-230000/rng_state_4.pth +3 -0
  26. checkpoint-230000/rng_state_5.pth +3 -0
  27. checkpoint-230000/rng_state_6.pth +3 -0
  28. checkpoint-230000/rng_state_7.pth +3 -0
  29. checkpoint-230000/scheduler.pt +3 -0
  30. checkpoint-230000/trainer_state.json +3 -0
  31. checkpoint-230000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -74,3 +74,5 @@ checkpoint-209000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
74
  checkpoint-210000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
75
  checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
74
  checkpoint-210000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
75
  checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
+ checkpoint-229000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
+ checkpoint-230000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-229000/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-229000/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-229000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15df79683599b550f42491d4ce469d3fb31a714a921564e0493e55891947f7a5
3
+ size 4411759432
checkpoint-229000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b0c02994c473318126d29e144f9a06bd318b858ac59d9143e531e1e3774b521
3
+ size 6332050591
checkpoint-229000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7cfdfb13f4188ffdfe7e4214372452214309978a2951f3cacf872ac8a9fdff4c
3
+ size 16389
checkpoint-229000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a08aeff7e006a727d9343e395f4f54f850b2ec3ed2984f79955d99deca8ea77
3
+ size 16389
checkpoint-229000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:158b62a3a8b492b6a585d436d8542c48abd230fc52c2ee7aff056c3deb9dbc91
3
+ size 16389
checkpoint-229000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52321d163952b114a5c8e2e391c49a72ea65cf3c1781ecfe4429079beeabe2a3
3
+ size 16389
checkpoint-229000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48abdc536ed9aceb049cd4bfa6945a527e3150d8c3fbc96dd3cf1015c0636afa
3
+ size 16389
checkpoint-229000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6a8c39a6e85332baad5aa3beec1ff16011e4233f29eeb3b0ae28a224047f7c4
3
+ size 16389
checkpoint-229000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f77372be462c20c359e156fc0d76ea45e412d2be2420452df1e3d3344a0606d0
3
+ size 16389
checkpoint-229000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f4fcec5a236238f49a7fd4b43fec85c8c6c3442db5b26449541407e1955ceb8
3
+ size 16389
checkpoint-229000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:acbfed64310863a7d9fa06891a116f6fc02d7c38650bfa4c26884ecdcc3cd817
3
+ size 1465
checkpoint-229000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0140823dd2ab3b26381a26cd63695b185ba9c98bcd4715eaea00f83e84814456
3
+ size 35241118
checkpoint-229000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17826cb02b5ee5534325fa853dcb15ab62a6320e3b6e6ed106d158cc19899229
3
+ size 6481
checkpoint-230000/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-230000/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-230000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fd4e7c3987b352454c73374642a44aef899948ef058b83e666a657241000b8f
3
+ size 4411759432
checkpoint-230000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a46b78d96d5759dda056baae923d54812de09bdf1b2b66b6a4d88811d1a50aa
3
+ size 6332050591
checkpoint-230000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6dc36eb3160e678e07805ef5d1021ea79f6e3e32ac70da1d5a0ab0bc53a8fd76
3
+ size 16389
checkpoint-230000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdeb1f58b33fa99f32c25a89c2ab1354df20e8ccfb793e4162841df39f0e6b55
3
+ size 16389
checkpoint-230000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:051bdb0853f6da668ecaab3143188c1f1603891d26db8d96fba90884d003dc73
3
+ size 16389
checkpoint-230000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73aec4f982b84c628f8d5054af25699db6c03c9145a7454b0b3c33d242808926
3
+ size 16389
checkpoint-230000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78264971aa8892822df28858e38d4bce631ab4c287e5dbbe0a3543765a55b2d7
3
+ size 16389
checkpoint-230000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf73d7b01de6316673b69527c80af568ce9a9161718f84087693e4e618b1fed8
3
+ size 16389
checkpoint-230000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:157943fba79510335673e45b385b4727dafbe8f038b4d7734c7332a98ca882f7
3
+ size 16389
checkpoint-230000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1202427a6df7637e25cc69989e6b8af60e14652c853cd5090d313e39f11aa6df
3
+ size 16389
checkpoint-230000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb323658c5e0f34d2ae6f6d245315fed80f6a7912726181af5e22a9863c5fc68
3
+ size 1465
checkpoint-230000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c349bea188f7f59fefd1a0d232bf92b525c1ae2bfd664843ca42d53baf9d82de
3
+ size 35395377
checkpoint-230000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:596e07cdca8b39cf1670c43ac94e64cd1fc233054937af74d3a591b0b36d0702
3
+ size 6417