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
4388d3f
·
verified ·
1 Parent(s): efcdc10

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-224000/config.json +79 -0
  3. checkpoint-224000/generation_config.json +6 -0
  4. checkpoint-224000/model.safetensors +3 -0
  5. checkpoint-224000/optimizer.pt +3 -0
  6. checkpoint-224000/rng_state_0.pth +3 -0
  7. checkpoint-224000/rng_state_1.pth +3 -0
  8. checkpoint-224000/rng_state_10.pth +3 -0
  9. checkpoint-224000/rng_state_11.pth +3 -0
  10. checkpoint-224000/rng_state_12.pth +3 -0
  11. checkpoint-224000/rng_state_13.pth +3 -0
  12. checkpoint-224000/rng_state_14.pth +3 -0
  13. checkpoint-224000/rng_state_15.pth +3 -0
  14. checkpoint-224000/rng_state_2.pth +3 -0
  15. checkpoint-224000/rng_state_3.pth +3 -0
  16. checkpoint-224000/rng_state_4.pth +3 -0
  17. checkpoint-224000/rng_state_5.pth +3 -0
  18. checkpoint-224000/rng_state_6.pth +3 -0
  19. checkpoint-224000/rng_state_7.pth +3 -0
  20. checkpoint-224000/rng_state_8.pth +3 -0
  21. checkpoint-224000/rng_state_9.pth +3 -0
  22. checkpoint-224000/scheduler.pt +3 -0
  23. checkpoint-224000/trainer_state.json +3 -0
  24. checkpoint-224000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -86,3 +86,4 @@ checkpoint-212000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
86
  checkpoint-218000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
87
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
88
  checkpoint-222000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
86
  checkpoint-218000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
87
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
88
  checkpoint-222000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
89
+ checkpoint-224000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-224000/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-224000/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-224000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f5f6af143d53aa11aacbf2617ad45efa6224605d54a588c337e7c8a2a7f4c58
3
+ size 4539487822
checkpoint-224000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42c4310d3b86823295258e4c6eca987df38d0226199bc8b79512fec8d348b1b6
3
+ size 6632345482
checkpoint-224000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96e62c651fef3a8872f4a403d110f8bd5b1fceec5cdb85a57ceebc44311fc5a9
3
+ size 15984
checkpoint-224000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed62eb9492b41efa40a3418286beda3144100e04138a0cbbb47de206430e02a6
3
+ size 15984
checkpoint-224000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:868882218c44d962da41a7a642b30a4fd257b4347613179c58e6f17d8c76b08b
3
+ size 15997
checkpoint-224000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30d733672c00645e052a9c9cbee305182e2f688fc2a7694cdd8c116049e5a7ae
3
+ size 15997
checkpoint-224000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5e99f2a94af485ae619dbfa14c0d9316eb1c06d8115b9ffc7bf14a7bed54963
3
+ size 15997
checkpoint-224000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57c9e2078fef3d43a30a267298f9298f9a55440aa008f509b6409f57f8aa3cf2
3
+ size 15997
checkpoint-224000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f8ab0cf076bca575d4c57d7685ef6bdcee30e45799ca32c65f589c771a271d3
3
+ size 15997
checkpoint-224000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d881427612e2d3b679ad18525685cf2654369cffda8203cf3ca3a4bacf4dca8
3
+ size 15997
checkpoint-224000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:389ddb2e67aef87da548081724a5cd1ef85ca71c9192d7df9a4834cfce76f9f0
3
+ size 15984
checkpoint-224000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63d346ef9ce853121ae16bcffb24f3b321c65d0f1f434d7f3ea9b173dda3194c
3
+ size 15984
checkpoint-224000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4814392b2b6c3f9f200812aae2959e78ab6ee9ecbc17fad00743461ea5db717e
3
+ size 15984
checkpoint-224000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:304f27d0699a37bb1a57853ef621b60a93389763721b00b981b975e2f40d58f2
3
+ size 15984
checkpoint-224000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63dd7df98e9f8351af5571b194c8e38330859d1cc1ccd833fca744aee0727c19
3
+ size 15984
checkpoint-224000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96ec6f1eae00963211901217600ae616c937d4fe842bd69e747eef89ed18a7c2
3
+ size 15984
checkpoint-224000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1679038eabf455940e9ed0ba0be92600e4c9291ecb398a4c80ebd089df83952
3
+ size 15984
checkpoint-224000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf7d91152bdee55e93d51d11e9eab0ae02efdf4c782e2a78796c13d38787d7ca
3
+ size 15984
checkpoint-224000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1417ddf005cee504878d6eaaedabbe9cb03c79c968946aac437fa62705178883
3
+ size 1064
checkpoint-224000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5c1825f155e30ba01166f58bcd863fbf070210757e59603c8880127ff6bd573
3
+ size 34635590
checkpoint-224000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23ca323149cadb538d88d81b4700d35ffddecf9f96d4e095d01ed42dd14e6095
3
+ size 6008