Hugging Face's logo

zhiyang1
/
vlm_dc-vae-f32c32-sana-1.1-768_epoch-64-0_group-14_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
b7fb555
·
verified ·
1 Parent(s): 8b17540

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-39500/config.json +79 -0
  2. checkpoint-39500/generation_config.json +6 -0
  3. checkpoint-39500/model.safetensors +3 -0
  4. checkpoint-39500/optimizer.pt +3 -0
  5. checkpoint-39500/rng_state_0.pth +3 -0
  6. checkpoint-39500/rng_state_1.pth +3 -0
  7. checkpoint-39500/rng_state_2.pth +3 -0
  8. checkpoint-39500/rng_state_3.pth +3 -0
  9. checkpoint-39500/rng_state_4.pth +3 -0
  10. checkpoint-39500/rng_state_5.pth +3 -0
  11. checkpoint-39500/rng_state_6.pth +3 -0
  12. checkpoint-39500/rng_state_7.pth +3 -0
  13. checkpoint-39500/scheduler.pt +3 -0
  14. checkpoint-39500/trainer_state.json +0 -0
  15. checkpoint-39500/training_args.bin +3 -0
  16. checkpoint-40000/config.json +79 -0
  17. checkpoint-40000/generation_config.json +6 -0
  18. checkpoint-40000/model.safetensors +3 -0
  19. checkpoint-40000/optimizer.pt +3 -0
  20. checkpoint-40000/rng_state_0.pth +3 -0
  21. checkpoint-40000/rng_state_1.pth +3 -0
  22. checkpoint-40000/rng_state_2.pth +3 -0
  23. checkpoint-40000/rng_state_3.pth +3 -0
  24. checkpoint-40000/rng_state_4.pth +3 -0
  25. checkpoint-40000/rng_state_5.pth +3 -0
  26. checkpoint-40000/rng_state_6.pth +3 -0
  27. checkpoint-40000/rng_state_7.pth +3 -0
  28. checkpoint-40000/scheduler.pt +3 -0
  29. checkpoint-40000/trainer_state.json +0 -0
  30. checkpoint-40000/training_args.bin +3 -0
checkpoint-39500/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": 14,
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-39500/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-39500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72930f426e829091e6128e4c2417a9a3cd6cf3e5a64cdab4a4c100e74f3e62aa
3
+ size 4539487822
checkpoint-39500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4495ab3c97a69523a87f41cfff51b67fd302683a7a0c406f25909a609f114204
3
+ size 6632479704
checkpoint-39500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34944bcd02da36a29c54845dea121bfd79dfd25093754a6947cc6c13cee11f6c
3
+ size 15984
checkpoint-39500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb83af7d14b1a329160732ef65c09d61430b176205d21450ba28fd61b2ae0f2e
3
+ size 15984
checkpoint-39500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:501337b78df3c1d36d9d7606cc62127a4fed4537436b1ed56613db3bd933dd50
3
+ size 15984
checkpoint-39500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e3851f85c7531075d556b558c7f92335246e7152164887f85e878d637a19c3f
3
+ size 15984
checkpoint-39500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c51bbf000c896635394bc7a0fd5aff28f1f58699a2f281861857340010f486fb
3
+ size 15984
checkpoint-39500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00ec41f7a234311c8e2909cab3fcd21f3a5f1021203062bd2d1e6c3497b42e77
3
+ size 15984
checkpoint-39500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa6ca7b91e1e522c5f2a9a464f772345b60a73c23ea51f454bff9cdfbdebf00d
3
+ size 15984
checkpoint-39500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3be7165f100fa88b28b654a41ddf4206a1bc78036b4cc1f5de10f70f6b81cf2b
3
+ size 15984
checkpoint-39500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2272d541c9dd2d06c2761a16d093c276c4469b8f9d4acb59ee605c62f8397d7b
3
+ size 1064
checkpoint-39500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fba0dc58f4912e66e5d328cf3ef162ba388820d42ee9dac84e4a3f59a06b340
3
+ size 6072
checkpoint-40000/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": 14,
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-40000/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-40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9c273090f26257477cf239dd3ebcb744659745f57ea708e1af5be2a8e88bcd2
3
+ size 4539487822
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfb27bec5a2d80d7f95f4d78dea677a67fda5463197ba39edea95db1f5f45c0f
3
+ size 6632479704
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fae28502d882ca5f4d1355b5281d5bf5b8bc4fafa6cbff1f04dcc115588e467d
3
+ size 15984
checkpoint-40000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec9f21fc90430143a0a15cea8eac752271a0a713ab1a50215a70f2472f6014fb
3
+ size 15984
checkpoint-40000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0dd8f5e9126a05da0a8fc6f2b708afceffe19109ba79ca7d859dd2fc4132a88f
3
+ size 15984
checkpoint-40000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46e33eb8f0b8778775e24f3da000034ce44c9ac61f5a8aec7b377f5b2728ef6f
3
+ size 15984
checkpoint-40000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40661afd3caccdb489c727586610a942339b0157d179fb27d7f0c596b64dfb4e
3
+ size 15984
checkpoint-40000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:655eb4e721f397c58a661e68de63bb28cf81746f084d56d49459c25a3ab7d78f
3
+ size 15984
checkpoint-40000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4381ffd52d8e8d3800eba0518203b6832e316d6a08c766ec3b7dc25e75206819
3
+ size 15984
checkpoint-40000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82deb411cc77b711c7ed18ab335ff57d1192831e2dee1090b4a9112f145b81fe
3
+ size 15984
checkpoint-40000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:825e1818d8a0f8348d506acdab26ab27594e76e01808880216f9807387891534
3
+ size 1064
checkpoint-40000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-40000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fba0dc58f4912e66e5d328cf3ef162ba388820d42ee9dac84e4a3f59a06b340
3
+ size 6072