Hugging Face's logo

zhiyang1
/
vlm_dc-ae-f32c32-sana-1.1-diffusers-768_patch-2_epoch-64-0_group-7_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
6b97099
·
verified ·
1 Parent(s): 146c91b

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-79500/config.json +79 -0
  3. checkpoint-79500/generation_config.json +6 -0
  4. checkpoint-79500/model.safetensors +3 -0
  5. checkpoint-79500/optimizer.pt +3 -0
  6. checkpoint-79500/rng_state_0.pth +3 -0
  7. checkpoint-79500/rng_state_1.pth +3 -0
  8. checkpoint-79500/rng_state_2.pth +3 -0
  9. checkpoint-79500/rng_state_3.pth +3 -0
  10. checkpoint-79500/rng_state_4.pth +3 -0
  11. checkpoint-79500/rng_state_5.pth +3 -0
  12. checkpoint-79500/rng_state_6.pth +3 -0
  13. checkpoint-79500/rng_state_7.pth +3 -0
  14. checkpoint-79500/scheduler.pt +3 -0
  15. checkpoint-79500/trainer_state.json +3 -0
  16. checkpoint-79500/training_args.bin +3 -0
  17. checkpoint-80000/config.json +79 -0
  18. checkpoint-80000/generation_config.json +6 -0
  19. checkpoint-80000/model.safetensors +3 -0
  20. checkpoint-80000/optimizer.pt +3 -0
  21. checkpoint-80000/rng_state_0.pth +3 -0
  22. checkpoint-80000/rng_state_1.pth +3 -0
  23. checkpoint-80000/rng_state_2.pth +3 -0
  24. checkpoint-80000/rng_state_3.pth +3 -0
  25. checkpoint-80000/rng_state_4.pth +3 -0
  26. checkpoint-80000/rng_state_5.pth +3 -0
  27. checkpoint-80000/rng_state_6.pth +3 -0
  28. checkpoint-80000/rng_state_7.pth +3 -0
  29. checkpoint-80000/scheduler.pt +3 -0
  30. checkpoint-80000/trainer_state.json +3 -0
  31. checkpoint-80000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -35,3 +35,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
+ checkpoint-79500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
+ checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-79500/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": 2,
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-sana-1.1-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-79500/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-79500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3132ec7c2b98e1a083f41154da02e3253d42250f147fc134b5a41e9c5fc60c1
3
+ size 4539487822
checkpoint-79500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea01c1b8c3627052341c3c953df8eed55e236ac22f1f036da688546aa68f6a87
3
+ size 6632345482
checkpoint-79500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb4988c52b5fff8b346daab50273173b63a0481f1b9a59d35cd928f7559d8aa4
3
+ size 15984
checkpoint-79500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c35e2a79968b0ce97a2a3289c81c4ec049fcb3c9a47d6689302d2896ae338b8
3
+ size 15984
checkpoint-79500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5fcc03db610dfb5d39f6cf2314e38551dc5b1200e1659624d822171b7e153c5
3
+ size 15984
checkpoint-79500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d317218a95e876f4ca8b633d483996975c44bba7b9f1cbe2599c7a99208dde8
3
+ size 15984
checkpoint-79500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:892e5dc539f91f1b6e582a1bbdb079b2aa4c900d904386bd848d2d0a1ea56656
3
+ size 15984
checkpoint-79500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08d20420bc5cd39d9b4a12a792a9a54e7c0246b451afc444b5652fd25a5148f3
3
+ size 15984
checkpoint-79500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02b666ccceeb1999476edd61b307a3d9570b17fbba87cc27781cd7026a80d41a
3
+ size 15984
checkpoint-79500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10e638a99ecac4640ea15e4a36bc2a600f17f41a0d64cdd3030f0f5dc2bf9150
3
+ size 15984
checkpoint-79500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28607f6ce1a9d4b8d54e0c3715d628d02d75128894d33111316daa0aa28adf9d
3
+ size 1064
checkpoint-79500/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:104ecc0f608a766515224b4ed834fd28bcff7b2d373ca984e0d8f8e0278ff5af
3
+ size 12104308
checkpoint-79500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82f75030c0b0f41eb483cdeaf1e51be7cbb9eaded4b3d164bb429c752fdd0e31
3
+ size 6072
checkpoint-80000/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": 2,
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-sana-1.1-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-80000/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-80000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc09e4358e85dfbf82b644cf8f74b4cd4f4828817de2af382cd99d49ae944b7e
3
+ size 4539487822
checkpoint-80000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a37169196689dc7c21b4f1154594c0849a5b38fc4690c6c371628352c7fde3c
3
+ size 6632345482
checkpoint-80000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58e76409ccf5130f60fb290c9b5d1a9ff7ab45ad9c32f7bd517974b8e95acfe8
3
+ size 15984
checkpoint-80000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe5aa348303f1325ecc51adf6d38d8ffd4bf4210f20f2849980d42af14c7701f
3
+ size 15984
checkpoint-80000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c6f9e1a3dd000461f77b8c11ae2043271e852c71f238139d66dc38a93020624
3
+ size 15984
checkpoint-80000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3db6ba3ebf16c124600f8ac6dd68183e83e26525ec9a6a1e0b981f424c22b33
3
+ size 15984
checkpoint-80000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:506b5b70946379be43cfc68c4ad4f082b01768e8ad5036d491b1711910f32dc2
3
+ size 15984
checkpoint-80000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503e02b5b3b5aa684886084566b2b28d46aab394cb44f0546da3c7f191302614
3
+ size 15984
checkpoint-80000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da5998c84c61bb5d1201e6c860431ae509faf16c13fa404fa4664e1bbd86ef56
3
+ size 15984
checkpoint-80000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b14987135a0f96bd01da899975d0b06e561968e22d92b7c1b8fbfd30825c3d54
3
+ size 15984
checkpoint-80000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28ba12d98196a8a695ec1ab1dd2645d9bcc955d9536dba006f419d3d477cfa2b
3
+ size 1064
checkpoint-80000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:733826c09b1fd3924735e0dde6b38ae516c6540e486e54afcde5697f2af3bdef
3
+ size 12180437
checkpoint-80000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82f75030c0b0f41eb483cdeaf1e51be7cbb9eaded4b3d164bb429c752fdd0e31
3
+ size 6072