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
a486ebf
·
verified ·
1 Parent(s): 7cf425c

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-249000/config.json +79 -0
  3. checkpoint-249000/generation_config.json +6 -0
  4. checkpoint-249000/model.safetensors +3 -0
  5. checkpoint-249000/optimizer.pt +3 -0
  6. checkpoint-249000/rng_state_0.pth +3 -0
  7. checkpoint-249000/rng_state_1.pth +3 -0
  8. checkpoint-249000/rng_state_2.pth +3 -0
  9. checkpoint-249000/rng_state_3.pth +3 -0
  10. checkpoint-249000/rng_state_4.pth +3 -0
  11. checkpoint-249000/rng_state_5.pth +3 -0
  12. checkpoint-249000/rng_state_6.pth +3 -0
  13. checkpoint-249000/rng_state_7.pth +3 -0
  14. checkpoint-249000/scheduler.pt +3 -0
  15. checkpoint-249000/trainer_state.json +3 -0
  16. checkpoint-249000/training_args.bin +3 -0
  17. checkpoint-250000/config.json +79 -0
  18. checkpoint-250000/generation_config.json +6 -0
  19. checkpoint-250000/model.safetensors +3 -0
  20. checkpoint-250000/optimizer.pt +3 -0
  21. checkpoint-250000/rng_state_0.pth +3 -0
  22. checkpoint-250000/rng_state_1.pth +3 -0
  23. checkpoint-250000/rng_state_2.pth +3 -0
  24. checkpoint-250000/rng_state_3.pth +3 -0
  25. checkpoint-250000/rng_state_4.pth +3 -0
  26. checkpoint-250000/rng_state_5.pth +3 -0
  27. checkpoint-250000/rng_state_6.pth +3 -0
  28. checkpoint-250000/rng_state_7.pth +3 -0
  29. checkpoint-250000/scheduler.pt +3 -0
  30. checkpoint-250000/trainer_state.json +3 -0
  31. checkpoint-250000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -78,3 +78,5 @@ 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
79
  checkpoint-239000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
78
  checkpoint-230000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
  checkpoint-239000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
+ checkpoint-249000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
+ checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-249000/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-249000/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-249000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f09f6cc85aa5eaf60b17fa362e2eb955b99d629fbf303ca041091c4f3f6f3db
3
+ size 4411759432
checkpoint-249000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:811e2edb4747560b441f61aea98ca291f0823bb0bde1b42dcb3703da502dbcba
3
+ size 6332050591
checkpoint-249000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41b91ba710bdf7e1b6bcd29151f7dd2f57f00563c52dc8a83a3fa4b45d9aa5f2
3
+ size 16389
checkpoint-249000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3f56b12177346a23e9316030b070c686eb563f5a62fa18c9db10fb148ea35f6
3
+ size 16389
checkpoint-249000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e0d4d0600748e3242640843e1e0060c25b07bc23b4ac26be5e3313da2c2b5d6
3
+ size 16389
checkpoint-249000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d5c9113f169063f14413bd08a7e168de892793570c12d10e5bbd33be631ccfb
3
+ size 16389
checkpoint-249000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa19bf03bce47ce25b2af30a0d326d6c71e74d9b9011d1a8c5ebba5fb43c1645
3
+ size 16389
checkpoint-249000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cebde435d4c064ed6c6990b35b3bcbf8ab4112c58da6e24c282244f869425d6e
3
+ size 16389
checkpoint-249000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f6b16c4375d717769c84a5826c074c09d251c86d75b76ebdfb5c276c87e1a77
3
+ size 16389
checkpoint-249000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0545cb91cf15916876430677a57fdb79431a3cd164844328ff168198a1ff4dda
3
+ size 16389
checkpoint-249000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de8e0b6fac58724186ac7bab5f9bef8163cee237a77f7652ac9c7f4455edd2fe
3
+ size 1465
checkpoint-249000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a1dc4ddecc62cc1e78442d6f91f09f334ba0028cc55327c07e644514e1e7eda
3
+ size 38328242
checkpoint-249000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a938a51ea3c3f4f786dc8c6bf043d886f31fa4b3774f41e0e7b9442bf9bb580
3
+ size 6481
checkpoint-250000/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-250000/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-250000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a607318a74c180dd938d0678277b61ec7fa67dc097d4f5e9014105693e22cc60
3
+ size 4411759432
checkpoint-250000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:198300083852f79d828117dca257d8a9fa210cdeb71dc9347612851f5023a389
3
+ size 6332050591
checkpoint-250000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd1248793feb52ba2a2a4144d379c26efe08bfda898319954c948bd2935c3247
3
+ size 16389
checkpoint-250000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fd1d460a81925e59743be4b31bf71c9e8748f298c1f3c6596973a0e1baddf65
3
+ size 16389
checkpoint-250000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7d37dcc710574599b4b4a3f2fd38f8ef3aa9b62ac863e70d9ba128fc7cd84f4
3
+ size 16389
checkpoint-250000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e5c00c5264869d8bd0ac6bfc83335349dc479bf5aa86c356437e58ecc95e1ab
3
+ size 16389
checkpoint-250000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d73e8bd566dd2f6a0a0b41ffa6591a0b6aaf889dc6c24c06fd2588056778aecd
3
+ size 16389
checkpoint-250000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:214f637a04deda12ca8c1d1a272497144668c89dfebee0676e18707732e4a735
3
+ size 16389
checkpoint-250000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a6994843fe13ce3e165e2f5cf61296c588704228fbed157f80f6f9e18855160
3
+ size 16389
checkpoint-250000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2b2fda03c677e040580bedee4996ef8b6f99df2eb668671b59e57276494add7
3
+ size 16389
checkpoint-250000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae08e52079eed5e1e38cc7f73a8366ec78214a3e0057caafe120230589c939cf
3
+ size 1465
checkpoint-250000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab496b0d4e0b9daf4ebb5738901e2780de4020780331401995fe102729a6df62
3
+ size 38482535
checkpoint-250000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a938a51ea3c3f4f786dc8c6bf043d886f31fa4b3774f41e0e7b9442bf9bb580
3
+ size 6481