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
50c031b
·
verified ·
1 Parent(s): 7690540

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-29500/config.json +79 -0
  2. checkpoint-29500/generation_config.json +6 -0
  3. checkpoint-29500/model.safetensors +3 -0
  4. checkpoint-29500/optimizer.pt +3 -0
  5. checkpoint-29500/rng_state_0.pth +3 -0
  6. checkpoint-29500/rng_state_1.pth +3 -0
  7. checkpoint-29500/rng_state_2.pth +3 -0
  8. checkpoint-29500/rng_state_3.pth +3 -0
  9. checkpoint-29500/rng_state_4.pth +3 -0
  10. checkpoint-29500/rng_state_5.pth +3 -0
  11. checkpoint-29500/rng_state_6.pth +3 -0
  12. checkpoint-29500/rng_state_7.pth +3 -0
  13. checkpoint-29500/scheduler.pt +3 -0
  14. checkpoint-29500/trainer_state.json +0 -0
  15. checkpoint-29500/training_args.bin +3 -0
  16. checkpoint-30000/config.json +79 -0
  17. checkpoint-30000/generation_config.json +6 -0
  18. checkpoint-30000/model.safetensors +3 -0
  19. checkpoint-30000/optimizer.pt +3 -0
  20. checkpoint-30000/rng_state_0.pth +3 -0
  21. checkpoint-30000/rng_state_1.pth +3 -0
  22. checkpoint-30000/rng_state_2.pth +3 -0
  23. checkpoint-30000/rng_state_3.pth +3 -0
  24. checkpoint-30000/rng_state_4.pth +3 -0
  25. checkpoint-30000/rng_state_5.pth +3 -0
  26. checkpoint-30000/rng_state_6.pth +3 -0
  27. checkpoint-30000/rng_state_7.pth +3 -0
  28. checkpoint-30000/scheduler.pt +3 -0
  29. checkpoint-30000/trainer_state.json +0 -0
  30. checkpoint-30000/training_args.bin +3 -0
checkpoint-29500/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-29500/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-29500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e58dc592e2fd4e626b2beb7a6fe5d4a2ad9f22c88330b2d1299de4d84b23ae4
3
+ size 4539487822
checkpoint-29500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c67c732640d863dda5162e2241012b3f754e826f1d0dec4c8449802137a0eec
3
+ size 6632345482
checkpoint-29500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ea6cd20e8fdfc0bb91db4e94a168442edc450e443303038d2a4b946b2367c35
3
+ size 15984
checkpoint-29500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26e89111ec23889dd283292a640a40fc1a9f0c8686285bc5fefbd17a6911f42f
3
+ size 15984
checkpoint-29500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef616580adb233dfa2c16f63c80bf747cd94be89b3fa59d333b6a300fe96d513
3
+ size 15984
checkpoint-29500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bae433bb2b7de2980d9e0e3f03e1b871927c740f7cfca186817a82e45b65b111
3
+ size 15984
checkpoint-29500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1381450209e4fa12fff327f8b488835c7480696aadda4624694a4c21a148aeaa
3
+ size 15984
checkpoint-29500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e77ff040363453b47d5c08600521c3a05c80c52c77419d70fe8f7b00a087afb1
3
+ size 15984
checkpoint-29500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c10930eefdc0d891c4c93c79c9da5aceb3639f843553f9bdff619b307653247
3
+ size 15984
checkpoint-29500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47f5847885acf63f0aaf29e4f020e56f791a4ed2e6e0fa5670551b08abbf754e
3
+ size 15984
checkpoint-29500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d1862f5c7e2f593151a9468c407722a0f45dcea5359fe8064e7d543ef1f9d23
3
+ size 1064
checkpoint-29500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-29500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0060c5f81a29acc28ff7d883ee457880139e0fb70fec69eeaa56aec14de2b796
3
+ size 6008
checkpoint-30000/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-30000/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-30000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc8171e636e75ec92fd836e9418faf2331c1a5e7895dfcc77f614c8e8137ef9e
3
+ size 4539487822
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d326deb8b828a88da8f7480f53ec35bec114449e99605622de0ea78168310b95
3
+ size 6632345482
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4eb1fcc9d38264e8bc3b4716c040b12e2dc52b80529ceb01fc8df959f6812040
3
+ size 15984
checkpoint-30000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a6208b6d5849d8654be9b1adf1ac15f60de52bfc30718f2404468f04740b8a1
3
+ size 15984
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0645cd5558e00afe2ca5a2bc10f9056491b1a033d99526fc59d8eadbd77e74b
3
+ size 15984
checkpoint-30000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc5f7218e2bfb9a0fbc918141f11aa0dc511f46c27b90a9cdb9520cdb6bdc3dd
3
+ size 15984
checkpoint-30000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b4bd5038e75128cad2c1c49126022ec92d999de32de0ddd5546cdc7fe2723f9
3
+ size 15984
checkpoint-30000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e15d1d39c22eb493489b96947094773b58673be9fb7de584571e66b247b0591a
3
+ size 15984
checkpoint-30000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b382785782417c8823be762c251b2fd7b5074c19ec9406916c3774fe93e512af
3
+ size 15984
checkpoint-30000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f88f7a22016cccd2c7643338587a102e6936ad5e0fd745265cbf888aac62afcc
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c776371f4bd2fd9cefbf39e502e7959eda8cd87e0af0a9780eda4f3fcd737144
3
+ size 1064
checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0060c5f81a29acc28ff7d883ee457880139e0fb70fec69eeaa56aec14de2b796
3
+ size 6008