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
d9fc0d4
·
verified ·
1 Parent(s): bf3e664

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-140000/config.json +79 -0
  3. checkpoint-140000/generation_config.json +6 -0
  4. checkpoint-140000/model.safetensors +3 -0
  5. checkpoint-140000/optimizer.pt +3 -0
  6. checkpoint-140000/rng_state_0.pth +3 -0
  7. checkpoint-140000/rng_state_1.pth +3 -0
  8. checkpoint-140000/rng_state_10.pth +3 -0
  9. checkpoint-140000/rng_state_11.pth +3 -0
  10. checkpoint-140000/rng_state_12.pth +3 -0
  11. checkpoint-140000/rng_state_13.pth +3 -0
  12. checkpoint-140000/rng_state_14.pth +3 -0
  13. checkpoint-140000/rng_state_15.pth +3 -0
  14. checkpoint-140000/rng_state_2.pth +3 -0
  15. checkpoint-140000/rng_state_3.pth +3 -0
  16. checkpoint-140000/rng_state_4.pth +3 -0
  17. checkpoint-140000/rng_state_5.pth +3 -0
  18. checkpoint-140000/rng_state_6.pth +3 -0
  19. checkpoint-140000/rng_state_7.pth +3 -0
  20. checkpoint-140000/rng_state_8.pth +3 -0
  21. checkpoint-140000/rng_state_9.pth +3 -0
  22. checkpoint-140000/scheduler.pt +3 -0
  23. checkpoint-140000/trainer_state.json +3 -0
  24. checkpoint-140000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -57,3 +57,4 @@ checkpoint-128000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-136000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
59
  checkpoint-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
57
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
  checkpoint-136000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
59
  checkpoint-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
60
+ checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-140000/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-140000/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-140000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77b32681a1b66dfce1daee58ed1f2dee8a6bcbcfc4133c90745f52ed59b6cad2
3
+ size 4539487822
checkpoint-140000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7738f9cda9b93b4f1f5f7cb0174181e0d5976711576175b3b6bf770dbda3c018
3
+ size 6632345482
checkpoint-140000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e32f9e5e7e576b50485296ccadf88a9bf62af79010121e60bcb2fa17f47e424
3
+ size 15984
checkpoint-140000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8de66d7593d7acf931c4fdbda614a8348fd5aae9880265087457f6590eb579a
3
+ size 15984
checkpoint-140000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2146b5c688df9f0fa2a8458c6e506cb9c3cd25c32b368bac39232106522ff7c
3
+ size 15997
checkpoint-140000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ea0d6d30b3ad688ebb45ff46cea9301a2c1668802d005717aedc00f51f6a21b
3
+ size 15997
checkpoint-140000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a59d8d91f9b41e4449dfac8ec7dee2adf77ac7abd6b5663915782af4a28b68c
3
+ size 15997
checkpoint-140000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06895c84bf95eb2dc2fd0ca14dda8e3572d8b4ce625a2d9eebf5c773def0ee9e
3
+ size 15997
checkpoint-140000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e78b84af7800fa49f61989bbde5fa20192daeaaea1b6dd4329760c9ff159d56
3
+ size 15997
checkpoint-140000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:424fcc2aad9854be4817585b0f58a23feacbb032c2f1a7fc32bb7b410b0b5a8c
3
+ size 15997
checkpoint-140000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b1ebfcd4968b48471a5bba93c8c338bd08d5dd3fd2c0f8f22a80a493d72965b
3
+ size 15984
checkpoint-140000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8238289949eb6636c339665dbe799e0ba9fe37698df4a141a7e39655c5be59d
3
+ size 15984
checkpoint-140000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ad8b1858128411ded9a7ce7a4842340bcd743be9d1415f000be93d654636805
3
+ size 15984
checkpoint-140000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11419cbc06b1083f2ac1d6e7c9a01bdc89e2334143e5313600a4c5c8a3bd105c
3
+ size 15984
checkpoint-140000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16f404dae9bf0048351eee0ac6ad0a8577f75f5b50706476a71e90180912103e
3
+ size 15984
checkpoint-140000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b318f1acff7e5930a7af0b501c1e8d858d6f4de35c5370a12b097a3ac8e4267b
3
+ size 15984
checkpoint-140000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:527a6a23ef6b51fc5e70c950acff37e0ea0ff7b211f2fbe228812f3c6f590f97
3
+ size 15984
checkpoint-140000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:862d0468088037f7ca0e17c928527a5fddf98f57c5c3bb0d26d507fafb234daa
3
+ size 15984
checkpoint-140000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba3f10af11b4450d1d39766d4216535b6f9f467c488fb2e5707c79acd3c9b465
3
+ size 1064
checkpoint-140000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa56dee00c48cb1d9544c04af9789f4b7fce218a0e1823a0514326f6f32d8349
3
+ size 21631316
checkpoint-140000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:048ee20e14103157c092a594b0d83d27e0edfcabd51d1300b92ec1d6325a77be
3
+ size 6008