Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
2e0cff9
·
verified ·
1 Parent(s): 1c61663

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-126000/config.json +79 -0
  3. checkpoint-126000/generation_config.json +6 -0
  4. checkpoint-126000/model.safetensors +3 -0
  5. checkpoint-126000/optimizer.pt +3 -0
  6. checkpoint-126000/rng_state_0.pth +3 -0
  7. checkpoint-126000/rng_state_1.pth +3 -0
  8. checkpoint-126000/rng_state_10.pth +3 -0
  9. checkpoint-126000/rng_state_11.pth +3 -0
  10. checkpoint-126000/rng_state_12.pth +3 -0
  11. checkpoint-126000/rng_state_13.pth +3 -0
  12. checkpoint-126000/rng_state_14.pth +3 -0
  13. checkpoint-126000/rng_state_15.pth +3 -0
  14. checkpoint-126000/rng_state_2.pth +3 -0
  15. checkpoint-126000/rng_state_3.pth +3 -0
  16. checkpoint-126000/rng_state_4.pth +3 -0
  17. checkpoint-126000/rng_state_5.pth +3 -0
  18. checkpoint-126000/rng_state_6.pth +3 -0
  19. checkpoint-126000/rng_state_7.pth +3 -0
  20. checkpoint-126000/rng_state_8.pth +3 -0
  21. checkpoint-126000/rng_state_9.pth +3 -0
  22. checkpoint-126000/scheduler.pt +3 -0
  23. checkpoint-126000/trainer_state.json +3 -0
  24. checkpoint-126000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -57,3 +57,4 @@ checkpoint-119000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-124000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
59
  checkpoint-125000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
57
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-124000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
59
  checkpoint-125000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
60
+ checkpoint-126000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-126000/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-126000/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-126000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2f47339d2bf3a1f289972fee7aae2278f2870a8bef73670166728670684225f
3
+ size 4539487822
checkpoint-126000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3a930181c15e20916d4694888d36fc73ee10de163ef06fcffd8fed7c9e33732
3
+ size 6632345482
checkpoint-126000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca59c523481d9eef0450df749e36a73b632cab0130704c68135a45f1c868eed3
3
+ size 15984
checkpoint-126000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f937023e71b24a035b9f8ce48e37a1b9945c85ae3221f5fc6eff325f7a530380
3
+ size 15984
checkpoint-126000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc27dfffa870b0cf12a7b958f9ec30bc13abc92a296f72f9001d30e48529d1b8
3
+ size 15997
checkpoint-126000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6e06337631358cba9b0d64389a993661e3af45aa8c7d4ceb9911b3abd866476
3
+ size 15997
checkpoint-126000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3e934df6088566a95c6d9591173f453501124a01920214db2b6a4c9c44a6fe9
3
+ size 15997
checkpoint-126000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5be20bdcc22284e9a4c17e00cd9b732ac134cf216a77dd2af0cdc6ebf02460d4
3
+ size 15997
checkpoint-126000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47d610a5b7bf8462f982cf45c9fa330fe9463f6739e1c4254bbda7e2e9ac6484
3
+ size 15997
checkpoint-126000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46528a5d9ab987c2c9e5b4158282eca746094d91d08af2c4084e38b0530195d8
3
+ size 15997
checkpoint-126000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5dd7d5519f08a0ae1c7bdcef0a601a418a3e65afd40a29ac9657d8eddd9b0e28
3
+ size 15984
checkpoint-126000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb1317f428df562f5d8b54b9c8428ba9dc7d86ef38855731ec277e0b5a94beb9
3
+ size 15984
checkpoint-126000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e688c4820e9c41a0893c07f21265fb075408a45d3db721577de7f48753ecdd3f
3
+ size 15984
checkpoint-126000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac5ff651c1a2977667fcadccf3ffc1f2de68bda4294b928071b2c3811f7911f5
3
+ size 15984
checkpoint-126000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f408355bbc0b3b7d1844bff773ff37a1935e5aa1a75e9bf15d50dc1729d57036
3
+ size 15984
checkpoint-126000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb5dc47a304babf2b0951f04e2691d3b20a30b98257237f4c6b88e96a89d53e2
3
+ size 15984
checkpoint-126000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4fbdb00777d6361976050ac2aa1da2c2e2edf2aba2d72c4dad1313fcb85b19ff
3
+ size 15984
checkpoint-126000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6364adc8cceaf28fa094045f152e6ff30746b2f0e284dc05c9c035f76c88ed06
3
+ size 15984
checkpoint-126000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d07b93e48423b3ac48eaf0fcbe0afd1b6efa578f83cd3636e3e7f8e4a4893e4
3
+ size 1064
checkpoint-126000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34adecf803eddcd85829cacc9fd7212f01dd791edc2e148bed2daad54880ab9d
3
+ size 19465238
checkpoint-126000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6035e6f7b9ef89dd2d0cfe68deb5423e90237ff951924f8134ae1492fac973d6
3
+ size 6008