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
429264f
·
verified ·
1 Parent(s): 48581b8

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-59500/config.json +79 -0
  2. checkpoint-59500/generation_config.json +6 -0
  3. checkpoint-59500/model.safetensors +3 -0
  4. checkpoint-59500/optimizer.pt +3 -0
  5. checkpoint-59500/rng_state_0.pth +3 -0
  6. checkpoint-59500/rng_state_1.pth +3 -0
  7. checkpoint-59500/rng_state_2.pth +3 -0
  8. checkpoint-59500/rng_state_3.pth +3 -0
  9. checkpoint-59500/rng_state_4.pth +3 -0
  10. checkpoint-59500/rng_state_5.pth +3 -0
  11. checkpoint-59500/rng_state_6.pth +3 -0
  12. checkpoint-59500/rng_state_7.pth +3 -0
  13. checkpoint-59500/scheduler.pt +3 -0
  14. checkpoint-59500/trainer_state.json +0 -0
  15. checkpoint-59500/training_args.bin +3 -0
  16. checkpoint-60000/config.json +79 -0
  17. checkpoint-60000/generation_config.json +6 -0
  18. checkpoint-60000/model.safetensors +3 -0
  19. checkpoint-60000/optimizer.pt +3 -0
  20. checkpoint-60000/rng_state_0.pth +3 -0
  21. checkpoint-60000/rng_state_1.pth +3 -0
  22. checkpoint-60000/rng_state_2.pth +3 -0
  23. checkpoint-60000/rng_state_3.pth +3 -0
  24. checkpoint-60000/rng_state_4.pth +3 -0
  25. checkpoint-60000/rng_state_5.pth +3 -0
  26. checkpoint-60000/rng_state_6.pth +3 -0
  27. checkpoint-60000/rng_state_7.pth +3 -0
  28. checkpoint-60000/scheduler.pt +3 -0
  29. checkpoint-60000/trainer_state.json +0 -0
  30. checkpoint-60000/training_args.bin +3 -0
checkpoint-59500/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-59500/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-59500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d95d9bc932c3409c613b5b383386428a2b10298d83335b7c1fa572b88c761e67
3
+ size 4539487822
checkpoint-59500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22c20f9ce8f85256413185b325eabeca087c62f6b86151e0f9344f2638ace60d
3
+ size 6632345482
checkpoint-59500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e067ecbb6d28081dfe9f8a6ec3f32299d0a3fca0e45c9ced336bd012ed102820
3
+ size 15984
checkpoint-59500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d364c9fc1c9ecb7461a5a74e534d61e46d28f993e289ee544d82e6b8ded73655
3
+ size 15984
checkpoint-59500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ad3f3aa01bb5e2a963bdc20206dc7241e37289bc2829c7547def9c49734289c
3
+ size 15984
checkpoint-59500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7804008496adc740080119ddbf326525b215ddd50dea18f694697116269bfba
3
+ size 15984
checkpoint-59500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffdbbe75f74ee9b8f3427246b7e2dafcb0729355153cad89e9726adc6c83844d
3
+ size 15984
checkpoint-59500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb814162e49acc2606d92d693467d9d58741eabca323695f43c04706c69b49a9
3
+ size 15984
checkpoint-59500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e149b0c756e3172a097426003b3b6a45deb43e2be8a18fc662846efb3b61b936
3
+ size 15984
checkpoint-59500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5696dcf582c52bf416dc9380795fc9283fbf80c3d67a9cf62604dc329f2702e
3
+ size 15984
checkpoint-59500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0637fb85fc7a6edb234838ba9b578f380f09bec4f8a1964c0813d7048df28fec
3
+ size 1064
checkpoint-59500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-59500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d60c305013dbd9702ae2fd33633c31fff38231b8de5241e9fd6c84b2fc029c41
3
+ size 6072
checkpoint-60000/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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21c131a76c8adb14cb4c531c81521fe324d4d0a57088f52140b430b9790e1d1d
3
+ size 4539487822
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca6a2554f6e17b207b8f41d3b08b0ead71e7c914e4f9daf3c1ae7cc9615641b7
3
+ size 6632345482
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00b3ff9bdbd8ed1f7e8583e37a5c62f72dd2a66f4146e9fa24af332a685211cc
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de711710667695050d2f914e1ebe4effc4e68062b0707d87de40c48bee7a9748
3
+ size 15984
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1dae4b72b9438ea594b8298d9679ab7f8dc8b8315d5070e59c0b7a7724c1d73
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c4d80857bf46db4bc86820f3422717baec92ffc67706aed2b820fc5a3543633
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21cfd8372f880dbd5fe58b0a071cd06c67e5b508d2cf7a1490f938282c8c6fc5
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d02cf85b9b68b3c71165ed680ca6bbdfbbedf1f86ab9ff1224b54b45cca40a3
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32faa2f39ee567fdf5705739e350d85487c1ff7ab2dc87ca7adf230f1aa60e12
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bff3b82c7ea2b93e5af2925acc212d14e3dc716201406d004a9bff85ce1ed4bb
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:079138e825244be1b8548d5aa66049e1f176621b60a839f892f5958c46aaca22
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d60c305013dbd9702ae2fd33633c31fff38231b8de5241e9fd6c84b2fc029c41
3
+ size 6072