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
edf9f00
·
verified ·
1 Parent(s): cd831e8

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-160000/config.json +79 -0
  3. checkpoint-160000/generation_config.json +6 -0
  4. checkpoint-160000/model.safetensors +3 -0
  5. checkpoint-160000/optimizer.pt +3 -0
  6. checkpoint-160000/rng_state_0.pth +3 -0
  7. checkpoint-160000/rng_state_1.pth +3 -0
  8. checkpoint-160000/rng_state_10.pth +3 -0
  9. checkpoint-160000/rng_state_11.pth +3 -0
  10. checkpoint-160000/rng_state_12.pth +3 -0
  11. checkpoint-160000/rng_state_13.pth +3 -0
  12. checkpoint-160000/rng_state_14.pth +3 -0
  13. checkpoint-160000/rng_state_15.pth +3 -0
  14. checkpoint-160000/rng_state_2.pth +3 -0
  15. checkpoint-160000/rng_state_3.pth +3 -0
  16. checkpoint-160000/rng_state_4.pth +3 -0
  17. checkpoint-160000/rng_state_5.pth +3 -0
  18. checkpoint-160000/rng_state_6.pth +3 -0
  19. checkpoint-160000/rng_state_7.pth +3 -0
  20. checkpoint-160000/rng_state_8.pth +3 -0
  21. checkpoint-160000/rng_state_9.pth +3 -0
  22. checkpoint-160000/scheduler.pt +3 -0
  23. checkpoint-160000/trainer_state.json +3 -0
  24. checkpoint-160000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -63,3 +63,4 @@ checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-156000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  checkpoint-158000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
63
  checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-156000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  checkpoint-158000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
66
+ checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-160000/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-160000/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-160000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1f0c98a4c38c0a5752c7feec4d6c401ff62fa8a0e0fe0a9f2e9662a0f4c3ef1
3
+ size 4539487822
checkpoint-160000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89d26869c832cd6c16ee49d925e3eb5b619b90991ddb7d42bb02756aa28facf6
3
+ size 6632345482
checkpoint-160000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7aa478f27a4eae28e2b2f879d96959de3c5547939c3f9b748f176c3784fc5b93
3
+ size 15984
checkpoint-160000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:948d307368f8396229c38592add1d6ed608971615119eaa1cbeab65d4444b648
3
+ size 15984
checkpoint-160000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f99ab6ca96e2774aa0885e463e33ecc5177342f12167103ee31183f67bb6688d
3
+ size 15997
checkpoint-160000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31c4a012b5617f7d49f2708d0da755f1b99b2bc608b9e80a07f48f318af357c6
3
+ size 15997
checkpoint-160000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ff1e3b280481f3515edcc196c6710c63cc4a842bd3f77653809764dca84430d
3
+ size 15997
checkpoint-160000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87c0672ea3b91154a03c3a37d4e0160a55903d39c8bb38f3d4ba3e9cbd7d6e3e
3
+ size 15997
checkpoint-160000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07957e1a53acee7e266001ea2f16d5a5a826f82875e75fa5853b2105b5be2607
3
+ size 15997
checkpoint-160000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aebb7f68c6f869099f71ecb2381322710a92bd293918509953b2f0d666e42de0
3
+ size 15997
checkpoint-160000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3f3d7bff4fc9309c0b6a398050310f1d30a0bee04ee655befe89a048abb63cc
3
+ size 15984
checkpoint-160000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8476f2a5dfdddccb84d3c33cff66b543286552ec0cd332659b63bc43327baa04
3
+ size 15984
checkpoint-160000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:387d7a6ca839a6aaed4f3819cfd53a454797e89897c1a589b2da56c21abebdf4
3
+ size 15984
checkpoint-160000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb74305153366ebadd0a246eccb881aace17d91b2259e71d386aac432e77cc92
3
+ size 15984
checkpoint-160000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7eded7024f0cb28d6b4afd5ee2e4c77faa2a4d1f53587df9b4a9d282b9115ce0
3
+ size 15984
checkpoint-160000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:975dcf524cfd3632727a658074c5820a6c1218201994ac0ca6eba6cf07d95425
3
+ size 15984
checkpoint-160000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a832521318813340ad5d8b67f6e633e3ae829a485a3037c94761d4e2634100d6
3
+ size 15984
checkpoint-160000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e2246f37771745c9c33599f1b29952db2ddb058c1e9e96c0a64d00788557d41
3
+ size 15984
checkpoint-160000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ac2b8c55d3fdf021e3e93dc337344834d29976ab064706df3d638988b57fc8a
3
+ size 1064
checkpoint-160000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a068f745ca3b37eddbb7f1fabfa6af5b264fb5d31c0d07e12a047a96e151dee
3
+ size 24736489
checkpoint-160000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00d017688bab378a09f64924432480ddd1f684996513d26a829aba25a225ff7d
3
+ size 6008