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
3e07cec
·
verified ·
1 Parent(s): 50c031b

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": 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-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:c7f810f79815c8f10781d4aa00d78f2c257baffbd338c630725a931233d58f8f
3
+ size 4539487822
checkpoint-39500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b1c8ac459a458462c104005e69593edd9d063e27c4a04fda509023ea835a40e
3
+ size 6632345482
checkpoint-39500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5760cf7a488967f257dfa0ed8995faae1dbe72127fae64645b71a31c7ac877d
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:cddeead738f50054e6f62d25dbdb6b2a4d2afa46f32ceddea804896bb60b6b6a
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 sha256:4a9aa3d8bf47f745231d412073b8a9b88f66092bc3acdcdd440ffae914de4ce6
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 sha256:65aefdb889a8dfd12bd8925305b3de37c980130a7f5296a5723ee93fdcf526ce
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:cc8d75e37a6659b691a20329f4accd11fc7279d3b2c2f3a8235b8dbe225f2609
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:fa730118592c8f30f7666765d6d165c1c9f3adb963b4993e063c109b03f5d72a
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:c3c13acd4fd03e98c9ea03be5658c032baf7fd45fffa48dca5462a14a468de0a
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:f2241f97eb8658f895a9a8e759f28ed80ed960603b5e94d0eb0ed1c246dd7fb0
3
+ size 15984
checkpoint-39500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdc43d32a7018e47c8ea6d47772d8b1c523989b2ffea83a39ef336ee8934a4f2
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 sha256:6ac0508d0ddaf4b502921d6fee819860024b4195a3c438fcc5403b6bd5bfe5c8
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": 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-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 sha256:0e99311f50736c1e5889f0224715e56d4a2508c85ed03dd27b7da3813887b283
3
+ size 4539487822
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:717f3850244349a5283106a9adf0e8ebb80e2a48f2a7443ca6fe07dbd028149b
3
+ size 6632345482
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc111357f55b215406d3c5fbd566146976315228f51721b5bfbf1e3654d9aeb9
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:ac6f10f5c8a98abc2e9fb1b8c60fc1f8c30204b05d670a8207bfaf332e560f37
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:a2e490fe56acca702dbcd616302acec1d4a99755abc2b8d24db7d8152f48e6eb
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:bf576a195ef93049257d99f057a966b5ae8405b1ef128abd6b1795a4d14ce26f
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 sha256:73ca7121e3e98b7c5e9a640dda25145fe71f9e7297bac15f66bda002ffb43e3c
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:f3ee8955f9115a05b7366f54608a325ac41468870fa7f85aff245350156015ab
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 sha256:3d1cd964b645d3fe25563d0c5135a9ac0eb510126069f6766bf361c679131334
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 sha256:46a53663b8282b6684c7374cab0e85a69b1acaf2980d554067a1c48788f6e189
3
+ size 15984
checkpoint-40000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3a07aef5327ff8cfca0212da09c4c70a2293cebf639372b7fc9faa992bdf0fe
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 sha256:6ac0508d0ddaf4b502921d6fee819860024b4195a3c438fcc5403b6bd5bfe5c8
3
+ size 6072