Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_aug_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
ffe4ae3
·
verified ·
1 Parent(s): a6cb87f

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-30000/config.json +79 -0
  2. checkpoint-30000/generation_config.json +6 -0
  3. checkpoint-30000/model.safetensors +3 -0
  4. checkpoint-30000/optimizer.pt +3 -0
  5. checkpoint-30000/rng_state_0.pth +3 -0
  6. checkpoint-30000/rng_state_1.pth +3 -0
  7. checkpoint-30000/rng_state_10.pth +3 -0
  8. checkpoint-30000/rng_state_11.pth +3 -0
  9. checkpoint-30000/rng_state_12.pth +3 -0
  10. checkpoint-30000/rng_state_13.pth +3 -0
  11. checkpoint-30000/rng_state_14.pth +3 -0
  12. checkpoint-30000/rng_state_15.pth +3 -0
  13. checkpoint-30000/rng_state_2.pth +3 -0
  14. checkpoint-30000/rng_state_3.pth +3 -0
  15. checkpoint-30000/rng_state_4.pth +3 -0
  16. checkpoint-30000/rng_state_5.pth +3 -0
  17. checkpoint-30000/rng_state_6.pth +3 -0
  18. checkpoint-30000/rng_state_7.pth +3 -0
  19. checkpoint-30000/rng_state_8.pth +3 -0
  20. checkpoint-30000/rng_state_9.pth +3 -0
  21. checkpoint-30000/scheduler.pt +3 -0
  22. checkpoint-30000/trainer_state.json +0 -0
  23. checkpoint-30000/training_args.bin +3 -0
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 sha256:7711c87b4c1ae9b12fbba16c16ccbcd1f26b3cf20842e2ef3d73263d40f5affe
3
+ size 4539487822
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:532803372d83bee49a823e0a93bb0f39b15f04a948a1f12a59abf6f7d791fe71
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 sha256:0ab199b5b7de478abc668b9173ca84fad5954a0bc0657936bb0ac58a1d6b1b92
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 sha256:63c072dcfaeef71207281b8a882b6ec6098ba1528a3d1da8c66f93fb8b76b6f3
3
+ size 15984
checkpoint-30000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61a3be8f671be308ca71bc66ccccec0ba8b5edfdf91025972cdccca055b447f4
3
+ size 15997
checkpoint-30000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26f228ec35e77937e70483b827eb5ef06d4e3bbbb922f9c1af280055769663f1
3
+ size 15997
checkpoint-30000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56da2bd82289e417cfbef0ef336524151b4eced539946ca15ea7a074a61cef41
3
+ size 15997
checkpoint-30000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d495aeeb73db52bbc6e139f84a41e6d5a2c74ee1c5b546c2461ca24bbb5c9a2c
3
+ size 15997
checkpoint-30000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47a0bc5cfa1a46090293dea84d3f4a7b6ef0163240d0b27c68c76da7a5d5d161
3
+ size 15997
checkpoint-30000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c87f93feced2de4ebf9c1f4e4b6200e3c7be5672ffef3bf6a1ecfff4dae5c0cf
3
+ size 15997
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a5bac0443c763f2adb9eca63aea1fe5717a2c61c950172657ae978091c10aad
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 sha256:85c4b29c944579c7eb07ffd87ccf16d12f3434bf9be7d63172d2c1f14d8e183f
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:e13284d0ef64564972e0655251f6bfe50db5cda40b42d688c612107a2afc296b
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 sha256:10049ab5d834e4ae84ebe085dbfcdb85b40ffc4a8708b62ba214c4c973a27b43
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 sha256:8c8f9bfe0d3289c4fbf5feb71a01ab65e36d4ce6f23cabe2faa2db5dd9f0978d
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:ebf4356f31a369b3a58a54c32522fee02da231740387df7e44d79242b7708cb5
3
+ size 15984
checkpoint-30000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3ad1b0d08935be2fb42e8a0be68f5fbe7e5ac33434f92255ae9c9b65255d5e7
3
+ size 15984
checkpoint-30000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cae61ebb892fbd3e03ede36af1eda32d79f6496f04236e5ab82e3d8e6b88097
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a4e7b2ed346955b5b87c3fd2ab508efd139e94fd0e7ddb084f3d83700b63e85
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:c700f00357d98c857792498e735638a9040230bcbe37dfa80d9c60e470463747
3
+ size 6008