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
b95f13f
·
verified ·
1 Parent(s): 27fffb3

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-204000/config.json +79 -0
  3. checkpoint-204000/generation_config.json +6 -0
  4. checkpoint-204000/model.safetensors +3 -0
  5. checkpoint-204000/optimizer.pt +3 -0
  6. checkpoint-204000/rng_state_0.pth +3 -0
  7. checkpoint-204000/rng_state_1.pth +3 -0
  8. checkpoint-204000/rng_state_10.pth +3 -0
  9. checkpoint-204000/rng_state_11.pth +3 -0
  10. checkpoint-204000/rng_state_12.pth +3 -0
  11. checkpoint-204000/rng_state_13.pth +3 -0
  12. checkpoint-204000/rng_state_14.pth +3 -0
  13. checkpoint-204000/rng_state_15.pth +3 -0
  14. checkpoint-204000/rng_state_2.pth +3 -0
  15. checkpoint-204000/rng_state_3.pth +3 -0
  16. checkpoint-204000/rng_state_4.pth +3 -0
  17. checkpoint-204000/rng_state_5.pth +3 -0
  18. checkpoint-204000/rng_state_6.pth +3 -0
  19. checkpoint-204000/rng_state_7.pth +3 -0
  20. checkpoint-204000/rng_state_8.pth +3 -0
  21. checkpoint-204000/rng_state_9.pth +3 -0
  22. checkpoint-204000/scheduler.pt +3 -0
  23. checkpoint-204000/trainer_state.json +3 -0
  24. checkpoint-204000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -79,3 +79,4 @@ checkpoint-194000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
  checkpoint-198000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-200000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-202000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
79
  checkpoint-198000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-200000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-202000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
+ checkpoint-204000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-204000/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-204000/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-204000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15fc630de7e5a53345abe74ee9fdf713c05db0deee58c6d339d879db748bb201
3
+ size 4539487822
checkpoint-204000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dac31087d4025e7223e0d2cedf251bb9d5b7092595716f08a828a9931e235475
3
+ size 6632345482
checkpoint-204000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa9f0e7fef36721a61cd218ed3d41ed72b630735a19ea20eff72d8d6fd84fbfb
3
+ size 15984
checkpoint-204000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6afa503841703e689db78df6a61c9bed370dec8dbd43545c90dabcab1b2d69d
3
+ size 15984
checkpoint-204000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd26d3d6ca4820ef9e8f7d6ed7ce412b326dda5d46029c61adcd37057340a04e
3
+ size 15997
checkpoint-204000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f96a081219e1298d0ec117fbed03d3e93e08498b54583a678c51e8233241be57
3
+ size 15997
checkpoint-204000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f225d3f6623c6b958c67def25c1d5d1c6a44db92a9f3ed44125aa9b6a8928a45
3
+ size 15997
checkpoint-204000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81218e1e5810c8218fe0ecbfe8ec88b794f9350e54c67550555965036fd69c8d
3
+ size 15997
checkpoint-204000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:347172b3c98480ff61e03ab4a7dcd422124f0a17a2f51d7d82234669f2c04bfb
3
+ size 15997
checkpoint-204000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14340878bdebb8529ceafbfe04c2013b933f9b5a47ce6ca6ba9294b5629a6165
3
+ size 15997
checkpoint-204000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb2de568a6581486a7479a293c1e37ae318c6448e8cb420a07b0d076cfa71bc7
3
+ size 15984
checkpoint-204000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1088d03038647e16d4319ba0975fceee1b8f3eb20eb4701682042a4e70e04211
3
+ size 15984
checkpoint-204000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37c3c7c9e62fafb20b8624b4d8d7cd24b159bf083921ee87a593bc8ce35ce789
3
+ size 15984
checkpoint-204000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d219c53a21526d1c210b9f813612f2111c521d5d66b76632a117078e5fa56360
3
+ size 15984
checkpoint-204000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:751200a234fb2188dff41a37508924731b694b743cd1c0fbc285dc2eb480f6a0
3
+ size 15984
checkpoint-204000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:454ffc83769411ebb40362e990a467c105a226c8f893b2ad03a20797f3fd38e8
3
+ size 15984
checkpoint-204000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4fb26a1c92c582fb7a14a0ee042f2256288d42ea23409506ffeef5d1c4731245
3
+ size 15984
checkpoint-204000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55ba40cb6b289e682667675cb808c56b3cfec42ceb1144bfcc2d2e664fdef296
3
+ size 15984
checkpoint-204000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ded0542d4dd749ca535eaebc9882abf24914119a3ec4d97695a14cd34fb4d29
3
+ size 1064
checkpoint-204000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad897acd7de7c643d629106ec428fb8f7bfb1a914509f73f9227046a21d6e2f8
3
+ size 31542854
checkpoint-204000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79cbb0f3a5379b91cf7642eb49dd3e07a84d824642ef81c93f6bd168f9db5c3e
3
+ size 6008