Hugging Face's logo

zhiyang1
/
blip3o_f32c32-sana-1.0-768_patch-1_group-7_fusion_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
9aa9931
·
verified ·
1 Parent(s): 7fcd253

Upload folder using huggingface_hub

Browse files
Files changed (27) hide show
  1. .gitattributes +2 -0
  2. checkpoint-148000/config.json +79 -0
  3. checkpoint-148000/generation_config.json +6 -0
  4. checkpoint-148000/model-00001-of-00002.safetensors +3 -0
  5. checkpoint-148000/model-00002-of-00002.safetensors +3 -0
  6. checkpoint-148000/model.safetensors.index.json +0 -0
  7. checkpoint-148000/optimizer.pt +3 -0
  8. checkpoint-148000/rng_state_0.pth +3 -0
  9. checkpoint-148000/rng_state_1.pth +3 -0
  10. checkpoint-148000/rng_state_2.pth +3 -0
  11. checkpoint-148000/rng_state_3.pth +3 -0
  12. checkpoint-148000/scheduler.pt +3 -0
  13. checkpoint-148000/trainer_state.json +3 -0
  14. checkpoint-148000/training_args.bin +3 -0
  15. checkpoint-150000/config.json +79 -0
  16. checkpoint-150000/generation_config.json +6 -0
  17. checkpoint-150000/model-00001-of-00002.safetensors +3 -0
  18. checkpoint-150000/model-00002-of-00002.safetensors +3 -0
  19. checkpoint-150000/model.safetensors.index.json +0 -0
  20. checkpoint-150000/optimizer.pt +3 -0
  21. checkpoint-150000/rng_state_0.pth +3 -0
  22. checkpoint-150000/rng_state_1.pth +3 -0
  23. checkpoint-150000/rng_state_2.pth +3 -0
  24. checkpoint-150000/rng_state_3.pth +3 -0
  25. checkpoint-150000/scheduler.pt +3 -0
  26. checkpoint-150000/trainer_state.json +3 -0
  27. checkpoint-150000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -74,3 +74,5 @@ checkpoint-138000/coco_steps-40_cfg-5/coco_8_LOOKS[[:space:]]LIKE[[:space:]]A[[:
74
  checkpoint-138000/coco_steps-40_cfg-5/coco_9_Three[[:space:]]people[[:space:]]are[[:space:]]walking[[:space:]]down[[:space:]]the[[:space:]]road[[:space:]]with[[:space:]]three[[:space:]]horses._1.png filter=lfs diff=lfs merge=lfs -text
75
  checkpoint-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
74
  checkpoint-138000/coco_steps-40_cfg-5/coco_9_Three[[:space:]]people[[:space:]]are[[:space:]]walking[[:space:]]down[[:space:]]the[[:space:]]road[[:space:]]with[[:space:]]three[[:space:]]horses._1.png filter=lfs diff=lfs merge=lfs -text
75
  checkpoint-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
+ checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
+ checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-148000/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-148000/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-148000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:028f2b1cf26938c06878a8a538c44b5878ee60b2c572938d548a99b128a8e9fa
3
+ size 4974360904
checkpoint-148000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d399ddf55dbb67fbc4d4e738076ab4209d54ab2099bf5aff6241c41930f9bd67
3
+ size 3234250886
checkpoint-148000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-148000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b588f03429441f6ccc49c1e56d3a34a49e81e7cc6be5a9468eb05658edb05a4
3
+ size 7455061885
checkpoint-148000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9a7398af360a956ff2b9626db2f46b5db665604a94d17c28f73fb12f0785ab7
3
+ size 15024
checkpoint-148000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dad55ed7648ab1d111328d0d238b779a990f8c7ed227167d40890ac238f34c47
3
+ size 15024
checkpoint-148000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a9551bd8d328ebe6fac611794302bbee9f2ee6f307216ede69a4c2f3870f595
3
+ size 15024
checkpoint-148000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8364d4a2d58aa6a80be844a150bf28987b29282f2954f5f4c31acaa51d85de5a
3
+ size 15024
checkpoint-148000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ede12b49b56707a56eb6cff6946ba010bfc21e4d6c9d40bd05227680119aabc
3
+ size 1064
checkpoint-148000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f462f1614c60dabfc80e408a2dc6ae5a1a123026e6c0e40b945b1d37f3a9e58
3
+ size 22786219
checkpoint-148000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da74882696e9e8dd34d16ca4d22f38a06acfab4a405369b2be1e051ee109b298
3
+ size 6008
checkpoint-150000/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-150000/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-150000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:975b1b75d4f8331d511069a92f464bc1455ec9f84234cd5561998929f0e066f1
3
+ size 4974360904
checkpoint-150000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0de9de0db05c6ae827de52953ade714ea4135d68a563cf955dce128dc1d6508e
3
+ size 3234250886
checkpoint-150000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-150000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2cbeaf1f9c662c6b1ba9bb09cac4a55d5116f103affa9a67fcd1aeaf030d3ef
3
+ size 7455061885
checkpoint-150000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4a8075e9f6ddb85b6e8c6665877b70165ef997d48b7d4b7bc8eda9badf23f6f
3
+ size 15024
checkpoint-150000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b9ac68e0796b52835c26136cdaba0a2c16298e3c39f262ef65f7c90f9147632
3
+ size 15024
checkpoint-150000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cef5adb13b8240d73954353d3eb5142236d972abaa648bba9f2b5b01b1ade7e
3
+ size 15024
checkpoint-150000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bf0886dff73e2a07df5b57755cdfc0d775eafb45b4019a1f40b943c72ba99fe
3
+ size 15024
checkpoint-150000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb8596898f44287ea3811aa1699eacd7316cc0974a138e75d6f4825669fe5dc1
3
+ size 1064
checkpoint-150000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc219fc8cf988badaf35e155c97425132f5c137f503ab70064977a485b847e43
3
+ size 23095053
checkpoint-150000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da74882696e9e8dd34d16ca4d22f38a06acfab4a405369b2be1e051ee109b298
3
+ size 6008