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
48581b8
·
verified ·
1 Parent(s): 3e07cec

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-49500/config.json +79 -0
  2. checkpoint-49500/generation_config.json +6 -0
  3. checkpoint-49500/model.safetensors +3 -0
  4. checkpoint-49500/optimizer.pt +3 -0
  5. checkpoint-49500/rng_state_0.pth +3 -0
  6. checkpoint-49500/rng_state_1.pth +3 -0
  7. checkpoint-49500/rng_state_2.pth +3 -0
  8. checkpoint-49500/rng_state_3.pth +3 -0
  9. checkpoint-49500/rng_state_4.pth +3 -0
  10. checkpoint-49500/rng_state_5.pth +3 -0
  11. checkpoint-49500/rng_state_6.pth +3 -0
  12. checkpoint-49500/rng_state_7.pth +3 -0
  13. checkpoint-49500/scheduler.pt +3 -0
  14. checkpoint-49500/trainer_state.json +0 -0
  15. checkpoint-49500/training_args.bin +3 -0
  16. checkpoint-50000/config.json +79 -0
  17. checkpoint-50000/generation_config.json +6 -0
  18. checkpoint-50000/model.safetensors +3 -0
  19. checkpoint-50000/optimizer.pt +3 -0
  20. checkpoint-50000/rng_state_0.pth +3 -0
  21. checkpoint-50000/rng_state_1.pth +3 -0
  22. checkpoint-50000/rng_state_2.pth +3 -0
  23. checkpoint-50000/rng_state_3.pth +3 -0
  24. checkpoint-50000/rng_state_4.pth +3 -0
  25. checkpoint-50000/rng_state_5.pth +3 -0
  26. checkpoint-50000/rng_state_6.pth +3 -0
  27. checkpoint-50000/rng_state_7.pth +3 -0
  28. checkpoint-50000/scheduler.pt +3 -0
  29. checkpoint-50000/trainer_state.json +0 -0
  30. checkpoint-50000/training_args.bin +3 -0
checkpoint-49500/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-49500/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-49500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6795def5af88ab2ec5cd198e7ceeb43eeb57f0af68e8e629f7afb50279e4247
3
+ size 4539487822
checkpoint-49500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd9697981f681e58e5786b692e739c9db21637bf38c7bcfd5adc00efd45587d8
3
+ size 6632345482
checkpoint-49500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c21c2664c7c1a831dd619f3b6e751ea16b22bdf207ee1396c5e60773a4cb807
3
+ size 15984
checkpoint-49500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa38acdaa65b23fb6254c8b7f125ee4265aeb6aba798748a40d82c38dcd61b0c
3
+ size 15984
checkpoint-49500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d37eb09fc2ce67ebb0ab2e2b1c04bcfe743230cf6c476e0c98adc34bee3cd0d
3
+ size 15984
checkpoint-49500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1eea10c66137d996c9b7d2bd0311d22ff6a67855f1ae0103742090dff4412fa
3
+ size 15984
checkpoint-49500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7911e6c66055efae799dc8290559893d4a607350f994fd1f7f42d393d70e2e14
3
+ size 15984
checkpoint-49500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:247d5ef7aef08d43b87cc4a24a008b12b998cb319bcb858159ae57196440e571
3
+ size 15984
checkpoint-49500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba7200e577f0e83d62403f50329020ec14b5015857cf615bbc3f272523d1e6a7
3
+ size 15984
checkpoint-49500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6482d04374588e21a99ef2bb9056d3a26c8641d1f17f9db931fcd323bfdca656
3
+ size 15984
checkpoint-49500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17f462d2644bb399b7466ebc8d7abb0c26e93ac8a321d8bf699205f8085bec96
3
+ size 1064
checkpoint-49500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-49500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3f2a18c208b6a0a57e94dfdb5ea7e4ab75791f524357d8ea8a664aa6abd9216
3
+ size 6072
checkpoint-50000/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-50000/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-50000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2712d8cd2e68aad49ca5377680c02093c74edece54538307bec30f747aa967ca
3
+ size 4539487822
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0bbbc72968905090dda1f6e7ee841f287986f1094a5ec1be78347f7eb0916494
3
+ size 6632345482
checkpoint-50000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7db9a0ecca248e040bf14214f5d824b647898a9ef8a3a25cfedff43283b7b387
3
+ size 15984
checkpoint-50000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9f3b0b0bbd2068e12d272c4f7b41a2c1bc1f48ff5a3abf7e7e0fe4e7d9de24b
3
+ size 15984
checkpoint-50000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26f6908e0ba24796321269ecc2434b6008a1a6fa094156430647a8acca5b6409
3
+ size 15984
checkpoint-50000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de10f502a755ca3c816396e5b58c24163da767f08cad639716b62fe6d8790dba
3
+ size 15984
checkpoint-50000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5859f1bbba8ef1ceed30fce6e068e12f05caae607cc6a8fcb37bb40641242693
3
+ size 15984
checkpoint-50000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46c014a65e85acc7fb45240d6dcda83b2aac3bd7cdcb1921b6f8626b37d3a5a5
3
+ size 15984
checkpoint-50000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc6907bd9c4f64e493072b119f7df6bb4338353f3beca9280facb53202d8bb42
3
+ size 15984
checkpoint-50000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97df8b1b1707ee54dc3b5ec028e3ffc6a355234013b96adab4bfecf48390c6e8
3
+ size 15984
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9477c015b5c34ee1be8d68df8cacc1cbcbadda0473d9d116004e7a1dee180dc8
3
+ size 1064
checkpoint-50000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3f2a18c208b6a0a57e94dfdb5ea7e4ab75791f524357d8ea8a664aa6abd9216
3
+ size 6072