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
6a4ba5f
·
verified ·
1 Parent(s): fae71c8

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-92000/config.json +79 -0
  3. checkpoint-92000/generation_config.json +6 -0
  4. checkpoint-92000/model.safetensors +3 -0
  5. checkpoint-92000/optimizer.pt +3 -0
  6. checkpoint-92000/rng_state_0.pth +3 -0
  7. checkpoint-92000/rng_state_1.pth +3 -0
  8. checkpoint-92000/rng_state_10.pth +3 -0
  9. checkpoint-92000/rng_state_11.pth +3 -0
  10. checkpoint-92000/rng_state_12.pth +3 -0
  11. checkpoint-92000/rng_state_13.pth +3 -0
  12. checkpoint-92000/rng_state_14.pth +3 -0
  13. checkpoint-92000/rng_state_15.pth +3 -0
  14. checkpoint-92000/rng_state_2.pth +3 -0
  15. checkpoint-92000/rng_state_3.pth +3 -0
  16. checkpoint-92000/rng_state_4.pth +3 -0
  17. checkpoint-92000/rng_state_5.pth +3 -0
  18. checkpoint-92000/rng_state_6.pth +3 -0
  19. checkpoint-92000/rng_state_7.pth +3 -0
  20. checkpoint-92000/rng_state_8.pth +3 -0
  21. checkpoint-92000/rng_state_9.pth +3 -0
  22. checkpoint-92000/scheduler.pt +3 -0
  23. checkpoint-92000/trainer_state.json +3 -0
  24. checkpoint-92000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -41,3 +41,4 @@ checkpoint-82000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
  checkpoint-84000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
41
  checkpoint-84000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
+ checkpoint-92000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-92000/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-92000/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-92000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b5821d98138e6cf0476e97c43f5fcca476e1e7e473100c634114fa485a4d75c
3
+ size 4539487822
checkpoint-92000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ee589ad3861703e849071a03fa21b6ca998018cc712e2177245fdf9c5ffcfc9
3
+ size 6632345482
checkpoint-92000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a11f9d978cbb334222ece22eee4ab98eb0391576f46ff54a3a1f492a5bded64f
3
+ size 15984
checkpoint-92000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:def4042366475c0cfb0632ac4c20f685395ab7d682322ee0925be129bf29edc6
3
+ size 15984
checkpoint-92000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d501ea49d0566acf8bbe458f2b5b57c1df2b3301f001e46ec6aab9cacde683db
3
+ size 15997
checkpoint-92000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b38ac5e5a1912b111f3d21e11932cac5d6ed5db9285c158fcbec2d61189105e7
3
+ size 15997
checkpoint-92000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddf258084ca93477efabef8233a54a6eb222c67c43a7c3891e3d42efe6e509af
3
+ size 15997
checkpoint-92000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36a152590c135d724cb7d0b7482c5e09be8a892a968e8b60ef2f96d122eeabc7
3
+ size 15997
checkpoint-92000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dcc043adce9a953c8321d80a1fe707247bdcd259721ae8ff3fdfae4ca145a60
3
+ size 15997
checkpoint-92000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9130cc4ca90fbcea0df7fd08aafaf9facecce4c296d3c8542edc7d573219d35b
3
+ size 15997
checkpoint-92000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c4177a80af821555723fe47009ce1739a10992478c6e80a6c3b1a560d80378f
3
+ size 15984
checkpoint-92000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5fa016a845006b072b886010aa5fe97a94083f9e7ba5c62190b77695dc340a3
3
+ size 15984
checkpoint-92000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65e64601a9b87cc757bb1923e46b1e42b20213902e62420fbee876d0932b185e
3
+ size 15984
checkpoint-92000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59de9a8828e3ffb721b57dece82e644b540b4af3639e44493624f83f08ea7a29
3
+ size 15984
checkpoint-92000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d89bd78afd6f5625638371bdbd6059e5fafdb3978a80b877536712ea588b592e
3
+ size 15984
checkpoint-92000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1988c2dc50ca0c149f795d0589a06d2dc43f9d03557ad38cded3ff52010fa330
3
+ size 15984
checkpoint-92000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e22043e8b868cdeeb6d0940a8c0481d44525ac91afae53ab2894b84ba5e4bac7
3
+ size 15984
checkpoint-92000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d3b73288a5156463064163c01610e66ec8ff92f83968e72728ac2468dcc9f77
3
+ size 15984
checkpoint-92000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0bba5487ee96c086102608e80a31a2da7fa75aa20730ce39d2548e84ad36b59d
3
+ size 1064
checkpoint-92000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90e8507be23caf71791519e1cda79dfba1b58cbd207fead5fa395f23991bba72
3
+ size 14189162
checkpoint-92000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb52c9b76d20c0ecbe659c65c42faec58885c14c6e049f7b91ccb5b56932fae7
3
+ size 6008