Hugging Face's logo

zhiyang1
/
dc-vae-f32c32-sana-1.0-768_patch-1_scratch_group-7_fusion

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
fbf2450
·
verified ·
1 Parent(s): 0fb475e

Upload folder using huggingface_hub

Browse files
Files changed (27) hide show
  1. .gitattributes +1 -0
  2. checkpoint-68000/config.json +79 -0
  3. checkpoint-68000/generation_config.json +6 -0
  4. checkpoint-68000/model-00001-of-00002.safetensors +3 -0
  5. checkpoint-68000/model-00002-of-00002.safetensors +3 -0
  6. checkpoint-68000/model.safetensors.index.json +0 -0
  7. checkpoint-68000/optimizer.pt +3 -0
  8. checkpoint-68000/rng_state_0.pth +3 -0
  9. checkpoint-68000/rng_state_1.pth +3 -0
  10. checkpoint-68000/rng_state_2.pth +3 -0
  11. checkpoint-68000/rng_state_3.pth +3 -0
  12. checkpoint-68000/scheduler.pt +3 -0
  13. checkpoint-68000/trainer_state.json +0 -0
  14. checkpoint-68000/training_args.bin +3 -0
  15. checkpoint-70000/config.json +79 -0
  16. checkpoint-70000/generation_config.json +6 -0
  17. checkpoint-70000/model-00001-of-00002.safetensors +3 -0
  18. checkpoint-70000/model-00002-of-00002.safetensors +3 -0
  19. checkpoint-70000/model.safetensors.index.json +0 -0
  20. checkpoint-70000/optimizer.pt +3 -0
  21. checkpoint-70000/rng_state_0.pth +3 -0
  22. checkpoint-70000/rng_state_1.pth +3 -0
  23. checkpoint-70000/rng_state_2.pth +3 -0
  24. checkpoint-70000/rng_state_3.pth +3 -0
  25. checkpoint-70000/scheduler.pt +3 -0
  26. checkpoint-70000/trainer_state.json +3 -0
  27. checkpoint-70000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -2521,3 +2521,4 @@ checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:
2521
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_50.png filter=lfs diff=lfs merge=lfs -text
2522
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_6.png filter=lfs diff=lfs merge=lfs -text
2523
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_8.png filter=lfs diff=lfs merge=lfs -text
 
 
2521
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_50.png filter=lfs diff=lfs merge=lfs -text
2522
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_6.png filter=lfs diff=lfs merge=lfs -text
2523
  checkpoint-48000/images_steps-40_cfg-5/imagenet_9_ostrich,[[:space:]]Struthio[[:space:]]camelus_8.png filter=lfs diff=lfs merge=lfs -text
2524
+ checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-68000/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": 1,
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": 16,
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": 0.5,
44
+ "repa_layers": null,
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": false,
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-68000/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-68000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd7bbbb339756579cdce7cc3d290068cd17f76bfb8bd5193440cb5967197081c
3
+ size 4973325456
checkpoint-68000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4c4d8f66ca9272740e1fbc634bd7308c625e606acf69ec3577e0f8a33c74370
3
+ size 3234250886
checkpoint-68000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-68000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21e29dc53438d61529b57fc536967190cbcb27207f52e722981f40039dd789a5
3
+ size 7453278409
checkpoint-68000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abc90324eb89d3b02c2e2549c7eecfb45db48834bcf6c3b16491740f8c2273d3
3
+ size 15024
checkpoint-68000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abd64e0014d849d4465348bd13b557ec1532c465c015634fa43f6fa3101bd52b
3
+ size 15024
checkpoint-68000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7ce05563ea6250ea9819d2d794624ea740b5cdbf08e01dfe7e2e5f50672c4d9
3
+ size 15024
checkpoint-68000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:482341e6df378ab00f86aea4fdc70c48358b54039f6aae65a4efb62c0e79bc96
3
+ size 15024
checkpoint-68000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:077f68a02a085369f88ca47183665a225422b142620816c10499fa62b80038b2
3
+ size 1064
checkpoint-68000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-68000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6af33cf7a86650223424fdb310b005e7f490a94f427d57abeaff87047a3f3004
3
+ size 6008
checkpoint-70000/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": 1,
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": 16,
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": 0.5,
44
+ "repa_layers": null,
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": false,
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-70000/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-70000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1af6c22cdb19f47def6ee98cc1d22f48b11d94d5be79d820dfeef05b3269046
3
+ size 4973325456
checkpoint-70000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5de57bb52df1c1165f52da280e7e7dc482cc29dc85e2720c7817dfd93a72a964
3
+ size 3234250886
checkpoint-70000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-70000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b71dfec979010681e41318f5620e746816c9270f1e11c61c3072e43f5ef6b4d
3
+ size 7453278409
checkpoint-70000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f45e312167205bda5f3739772d6a6651442bcae004089b4cebc4b09a9f0ef0f
3
+ size 15024
checkpoint-70000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63ba460227a860e4aad212357eba39cf67e68c1391fa8230bff1162f591b9ac0
3
+ size 15024
checkpoint-70000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a98f1e0f7f68147ff046294749a7d4d1033d4a69f6cd9b4d4fd3d89371c5d719
3
+ size 15024
checkpoint-70000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9fa730dd44786b61e164df456df58e4b73da416db9dcb56bc40fde6c580521d
3
+ size 15024
checkpoint-70000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22c0159ab07cbe040d7cc75a9072cd9ef60ff049a9e3854ea4629192f4fd66a6
3
+ size 1064
checkpoint-70000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22da2ca42eafa7faf73653ad549fdb6ba20637e136f9dec0a078e3c1279a8a40
3
+ size 10760626
checkpoint-70000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6af33cf7a86650223424fdb310b005e7f490a94f427d57abeaff87047a3f3004
3
+ size 6008