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
2c9183b
·
verified ·
1 Parent(s): fbfeee9

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-279000/config.json +79 -0
  3. checkpoint-279000/generation_config.json +6 -0
  4. checkpoint-279000/model.safetensors +3 -0
  5. checkpoint-279000/optimizer.pt +3 -0
  6. checkpoint-279000/rng_state_0.pth +3 -0
  7. checkpoint-279000/rng_state_1.pth +3 -0
  8. checkpoint-279000/rng_state_2.pth +3 -0
  9. checkpoint-279000/rng_state_3.pth +3 -0
  10. checkpoint-279000/rng_state_4.pth +3 -0
  11. checkpoint-279000/rng_state_5.pth +3 -0
  12. checkpoint-279000/rng_state_6.pth +3 -0
  13. checkpoint-279000/rng_state_7.pth +3 -0
  14. checkpoint-279000/scheduler.pt +3 -0
  15. checkpoint-279000/trainer_state.json +3 -0
  16. checkpoint-279000/training_args.bin +3 -0
  17. checkpoint-280000/config.json +79 -0
  18. checkpoint-280000/generation_config.json +6 -0
  19. checkpoint-280000/model.safetensors +3 -0
  20. checkpoint-280000/optimizer.pt +3 -0
  21. checkpoint-280000/rng_state_0.pth +3 -0
  22. checkpoint-280000/rng_state_1.pth +3 -0
  23. checkpoint-280000/rng_state_2.pth +3 -0
  24. checkpoint-280000/rng_state_3.pth +3 -0
  25. checkpoint-280000/rng_state_4.pth +3 -0
  26. checkpoint-280000/rng_state_5.pth +3 -0
  27. checkpoint-280000/rng_state_6.pth +3 -0
  28. checkpoint-280000/rng_state_7.pth +3 -0
  29. checkpoint-280000/scheduler.pt +3 -0
  30. checkpoint-280000/trainer_state.json +3 -0
  31. checkpoint-280000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -84,3 +84,5 @@ checkpoint-259000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
84
  checkpoint-260000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
85
  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
 
 
 
84
  checkpoint-260000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
85
  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
checkpoint-279000/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-279000/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-279000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6cb777ad74d0b614cab2e9f284774d0d2a4fd85249f727439cd5ed0eeb3a78d
3
+ size 4411759432
checkpoint-279000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b57bd1662b44fadb08b331580e2dce7f60263232f0205cfc1beabe7eab4da751
3
+ size 6332050591
checkpoint-279000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f9b762594eb1b2ccf5e14055c25a14d536dc58209ce98d463fac40988aaaa88
3
+ size 16389
checkpoint-279000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:013c03c188062752974abeab39e186042d4bfaa37f05eae3760e7e2f5fad434f
3
+ size 16389
checkpoint-279000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503ecf389a0bb37be932313ecb5a80da6da6619f6a343d94d5667416d055d1c9
3
+ size 16389
checkpoint-279000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9893b6c0914feda076f00a24fd59cd9225f0e158ca5160ca5fad4aecb632c6fa
3
+ size 16389
checkpoint-279000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f5472f3edc5baf9708c9449ed3f274d20e13c48aedc8583c0273dd54b3efcbb
3
+ size 16389
checkpoint-279000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c800b3aee769e885fdec827819f12d5548422d101ffb2d2a0f86d5eb62e6f0cf
3
+ size 16389
checkpoint-279000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b213e3c67ed78f3b935749e23e5beac0ae99049a1c7dee795ca36d00973c6c27
3
+ size 16389
checkpoint-279000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87156c8ec82ed34f69ce9574230b145eb37311067f732cb58dfbc9b93da4dd0f
3
+ size 16389
checkpoint-279000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d87b150c25e1d384809f2ecf9a12df1b9ee7cefccb28c7b043ee49aabe440be
3
+ size 1465
checkpoint-279000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:153cf5f0593e31df86991770f4fdfc501b4d988ec1a222db687f00756c7dc52d
3
+ size 42958053
checkpoint-279000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:404ababfb2ed3e94196d0d188d06acf43200501436d86c15aa153c6a330bcfc3
3
+ size 6481
checkpoint-280000/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-280000/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-280000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b7b4c04d132869149c92a6a4174419966faaaa3ddfcc17ae3dc9b17fb60c424
3
+ size 4411759432
checkpoint-280000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aeae4c2c5dd1fb3b653a15d85e79b57796fb340295ac018e2e99851d39f0586c
3
+ size 6332050591
checkpoint-280000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b5a326211baef0d5f9aee0aa9e0ad4053d58dbeaad2a9766ea1122e60744d61
3
+ size 16389
checkpoint-280000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edd76e2ca1a6b44ac68be6088f6ff58168325605dba8767134f3b9de8fffa424
3
+ size 16389
checkpoint-280000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:152f991815f4a9aa4c85f92e55d78345f9ef1df04f32a4dfde794214480f7315
3
+ size 16389
checkpoint-280000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61c864f2353584bdc40c903902c484f74004da95c80b1a102148137eb21e90e8
3
+ size 16389
checkpoint-280000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e81bd632816f073530ea23a69cab9162cd21899b2209a9906b2244b5b84a83c9
3
+ size 16389
checkpoint-280000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57ee2ad8763b3ad13ee9dd9aee68e3f479f7ecebc5aad9d045ca36808c0a3f4f
3
+ size 16389
checkpoint-280000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ea52a49130740de27599b3f9ad11ba2e5675f2d1d1edfae2f557adce24dae3f
3
+ size 16389
checkpoint-280000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d8e8c4e7925a75368efeffc43ab219ac14afb57943e5dad50204aa166993088
3
+ size 16389
checkpoint-280000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a781488892515deeece40fc1e2506e16bb23b79cf2e65d02f3d3aa6b3be44dd1
3
+ size 1465
checkpoint-280000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:265efc4450866a83922f5e4388c25721af1057c3bc3ae57cb534da1695223268
3
+ size 43112377
checkpoint-280000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:404ababfb2ed3e94196d0d188d06acf43200501436d86c15aa153c6a330bcfc3
3
+ size 6481