Hugging Face's logo

zhiyang1
/
dc-vae-f32c32-sana-1.0-768_patch-1_epoch-64_group-7_fusion_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
a3e77d4
·
verified ·
1 Parent(s): 2f1b211

Upload folder using huggingface_hub

Browse files
Files changed (34) hide show
  1. checkpoint-19500/config.json +79 -0
  2. checkpoint-19500/generation_config.json +6 -0
  3. checkpoint-19500/model-00001-of-00002.safetensors +3 -0
  4. checkpoint-19500/model-00002-of-00002.safetensors +3 -0
  5. checkpoint-19500/model.safetensors.index.json +0 -0
  6. checkpoint-19500/optimizer.pt +3 -0
  7. checkpoint-19500/rng_state_0.pth +3 -0
  8. checkpoint-19500/rng_state_1.pth +3 -0
  9. checkpoint-19500/rng_state_2.pth +3 -0
  10. checkpoint-19500/rng_state_3.pth +3 -0
  11. checkpoint-19500/rng_state_4.pth +3 -0
  12. checkpoint-19500/rng_state_5.pth +3 -0
  13. checkpoint-19500/rng_state_6.pth +3 -0
  14. checkpoint-19500/rng_state_7.pth +3 -0
  15. checkpoint-19500/scheduler.pt +3 -0
  16. checkpoint-19500/trainer_state.json +0 -0
  17. checkpoint-19500/training_args.bin +3 -0
  18. checkpoint-20000/config.json +79 -0
  19. checkpoint-20000/generation_config.json +6 -0
  20. checkpoint-20000/model-00001-of-00002.safetensors +3 -0
  21. checkpoint-20000/model-00002-of-00002.safetensors +3 -0
  22. checkpoint-20000/model.safetensors.index.json +0 -0
  23. checkpoint-20000/optimizer.pt +3 -0
  24. checkpoint-20000/rng_state_0.pth +3 -0
  25. checkpoint-20000/rng_state_1.pth +3 -0
  26. checkpoint-20000/rng_state_2.pth +3 -0
  27. checkpoint-20000/rng_state_3.pth +3 -0
  28. checkpoint-20000/rng_state_4.pth +3 -0
  29. checkpoint-20000/rng_state_5.pth +3 -0
  30. checkpoint-20000/rng_state_6.pth +3 -0
  31. checkpoint-20000/rng_state_7.pth +3 -0
  32. checkpoint-20000/scheduler.pt +3 -0
  33. checkpoint-20000/trainer_state.json +0 -0
  34. checkpoint-20000/training_args.bin +3 -0
checkpoint-19500/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": 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-19500/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-19500/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5ad362f8254ec7938dcd0a59073db1a9dc1e606b5936511f5dff1afffb67fa3
3
+ size 4981660336
checkpoint-19500/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f12c4a4d8cf4d9fd375e44f79086e0500c42c8c2766341ca12ff14b07a9c50fb
3
+ size 1896367718
checkpoint-19500/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-19500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9bdc06b67ab516ee3e0a0f1c6e563b7de150ee96bcffa68d5cdcf08e8244508
3
+ size 7455061885
checkpoint-19500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:885e4431fdb97475183f29cd8b15d9c6d3230a3a638a4816ca3763c3686796ec
3
+ size 15984
checkpoint-19500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:693fb4feb74d2e78689fcd695a0986bb6f2ad07ab31e629e2ca273865efb579c
3
+ size 15984
checkpoint-19500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25ae7bc90e7475c573f758399de097f9c3032ecb5eb76238882c99bce17c9f49
3
+ size 15984
checkpoint-19500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c91985daaa2f5c74e0b947a019d6b06bbda7a28bf853135f3f0eb3d50be6020
3
+ size 15984
checkpoint-19500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45d592e92af139a11fa58db7fd204911bec17245dd83f07468a8eef62ee4c56a
3
+ size 15984
checkpoint-19500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27c0e14a3e063cc510dd48f800f9a845a15ae46b7bb011ef59b9c5d311d2e02c
3
+ size 15984
checkpoint-19500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de09498034efe55aead384bb1cb49a423148b726a4a6308707ea09e2a06bca2c
3
+ size 15984
checkpoint-19500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29c6ee085188c8108a80c4ca715728f5fa30fbcb2ffb7e0a3aa6bea9950e1c9a
3
+ size 15984
checkpoint-19500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8c4f4356bf542da20add0a095d0be55bc56c89026280d08edbcb6cd78f335e6
3
+ size 1064
checkpoint-19500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-19500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e32e3ce97f63cbd3e95fb0999884e9051b00bbf523f05f51d6051428e5b2c79
3
+ size 6008
checkpoint-20000/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": 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-20000/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-20000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77833a9ccaf60bacee76688fa3b5244974c93d8572a27ba8962449cac2206231
3
+ size 4981660336
checkpoint-20000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f12c4a4d8cf4d9fd375e44f79086e0500c42c8c2766341ca12ff14b07a9c50fb
3
+ size 1896367718
checkpoint-20000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94543cb4ce9ac26d170d6f77f33d07e1b776b330ea4d2b95c1142538802bdf03
3
+ size 7455061885
checkpoint-20000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0155aac6541531b84f49029f84b27480a8808583c6be7eab76875b2436527132
3
+ size 15984
checkpoint-20000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1658773682eaf513efc0aa822e097bf6fa5e61f5f2e8a20095ee223e2ecdb693
3
+ size 15984
checkpoint-20000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0aac3e6c56a80c31747c81f10f97b113b34f2f6b6fddd0be569841281cb1203
3
+ size 15984
checkpoint-20000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc9f4fccf203a03cca623e671acba5652daa77a09c74e0cbc6f5e5e3a3ec0920
3
+ size 15984
checkpoint-20000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91ebce52b75e2b578a94f179fae61f020720c96e07b5db5a6cbb21ea2fba3fc1
3
+ size 15984
checkpoint-20000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:460cc91cddd588a149373b499f0fb80d672342139294213dbe4c97166da7888e
3
+ size 15984
checkpoint-20000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aa3ff19dd93197263f36cff26e4c4c8fb310f054b7830b1a922e88060ef66ca
3
+ size 15984
checkpoint-20000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:218762cc3cc1b856671898fce2ad12341b2d47f604c45bbd470690e7fb7b29ad
3
+ size 15984
checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e2e11c20cafdc07bdb97287eb14f1d8b64ee1da506824356f200ecd35984722
3
+ size 1064
checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e32e3ce97f63cbd3e95fb0999884e9051b00bbf523f05f51d6051428e5b2c79
3
+ size 6008