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
755ba88
·
verified ·
1 Parent(s): 2ccc710

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-82000/config.json +79 -0
  3. checkpoint-82000/generation_config.json +6 -0
  4. checkpoint-82000/model.safetensors +3 -0
  5. checkpoint-82000/optimizer.pt +3 -0
  6. checkpoint-82000/rng_state_0.pth +3 -0
  7. checkpoint-82000/rng_state_1.pth +3 -0
  8. checkpoint-82000/rng_state_10.pth +3 -0
  9. checkpoint-82000/rng_state_11.pth +3 -0
  10. checkpoint-82000/rng_state_12.pth +3 -0
  11. checkpoint-82000/rng_state_13.pth +3 -0
  12. checkpoint-82000/rng_state_14.pth +3 -0
  13. checkpoint-82000/rng_state_15.pth +3 -0
  14. checkpoint-82000/rng_state_2.pth +3 -0
  15. checkpoint-82000/rng_state_3.pth +3 -0
  16. checkpoint-82000/rng_state_4.pth +3 -0
  17. checkpoint-82000/rng_state_5.pth +3 -0
  18. checkpoint-82000/rng_state_6.pth +3 -0
  19. checkpoint-82000/rng_state_7.pth +3 -0
  20. checkpoint-82000/rng_state_8.pth +3 -0
  21. checkpoint-82000/rng_state_9.pth +3 -0
  22. checkpoint-82000/scheduler.pt +3 -0
  23. checkpoint-82000/trainer_state.json +3 -0
  24. checkpoint-82000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -37,3 +37,4 @@ checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-76000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
  checkpoint-78000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
37
  checkpoint-76000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
  checkpoint-78000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
40
+ checkpoint-82000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-82000/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-82000/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-82000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:496fac6c1137ca9709188684d217e82f37d09cf8de02d59f581432425dfb9c72
3
+ size 4539487822
checkpoint-82000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60cd6bea69034ddc4c916888efc2d59c49ab506d464be660eede99212a6968e5
3
+ size 6632345482
checkpoint-82000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb84ed2f126a3cd6eea620254d1674c17a2adf29ef2a1572f88c64b5b99f359d
3
+ size 15984
checkpoint-82000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:819b71dd5c0a2a618693c6f113a4304413ad87032338896ce8d317cf54708f29
3
+ size 15984
checkpoint-82000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f5f88d64903e29eab88064837e553c38906795ac8728ff4f64f4e121e5b6edb
3
+ size 15997
checkpoint-82000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:981a688217144a69ae7c2f75590f105250a9f6b4f49cbda10e194c78544b6d44
3
+ size 15997
checkpoint-82000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d053121bd0a9b0bfa32113d626b50bcfb3cd413339f375d418e6dc49870ff6b0
3
+ size 15997
checkpoint-82000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06db249afe41724ad26b739a679a5191f20471fa1a890052fd4a820b55f5d89d
3
+ size 15997
checkpoint-82000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a15ecc3a9fdaead62ddd12e6c96e1792c95db359c390861a83bf587fb62bf20
3
+ size 15997
checkpoint-82000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e29ed01653fb00fb5bb6be06c3dc88fc90fbcea21d01986dafe1e5c3f83608be
3
+ size 15997
checkpoint-82000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c89dbe3a1848f3cf28144023c2023691f8c93d291bd2314171efffc71ee59d88
3
+ size 15984
checkpoint-82000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5600aa048e77322e470c79e09d524896e282b8df8776052a92442e476c9867ba
3
+ size 15984
checkpoint-82000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:627061776b20e6e86a6280cc8d02b494591b977d0c32d3f7e90b064db36d4d2b
3
+ size 15984
checkpoint-82000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f19b3d51d93262c815011e8d09628bb963340caf5dee27de6d234b9b836c7011
3
+ size 15984
checkpoint-82000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b60ca80a454da179819cafb029d070b49b5b173b7ed21b1ff54e5bceb8e9c055
3
+ size 15984
checkpoint-82000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e04039697073d7e7ad59573ce54bee6f25802175535766f7b64a093cd49a2067
3
+ size 15984
checkpoint-82000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e58f2f878f4b72482719ea15950a2ab6581dd6543f5ffd396f19550cbff12846
3
+ size 15984
checkpoint-82000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a38e6ada214ed3fcdd5af9ce990b5beb848db27508c0aff1aaef1ca93caf9651
3
+ size 15984
checkpoint-82000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c9baec1e57060a66c74b341261b3c9f50996621eea355dddbfc07e1aeff37cd
3
+ size 1064
checkpoint-82000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf4f08de5f4d33f77c9b1e7d1655d52f5f873b7d1f2d056b8991d67d8002dc3a
3
+ size 12646340
checkpoint-82000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aeec56ad2c8bc964fd26381965541950f67ad9f3100490fcf8d2d715acd431fb
3
+ size 6008