Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
ee9e49e
·
verified ·
1 Parent(s): 9c337df

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-113000/config.json +79 -0
  3. checkpoint-113000/generation_config.json +6 -0
  4. checkpoint-113000/model.safetensors +3 -0
  5. checkpoint-113000/optimizer.pt +3 -0
  6. checkpoint-113000/rng_state_0.pth +3 -0
  7. checkpoint-113000/rng_state_1.pth +3 -0
  8. checkpoint-113000/rng_state_10.pth +3 -0
  9. checkpoint-113000/rng_state_11.pth +3 -0
  10. checkpoint-113000/rng_state_12.pth +3 -0
  11. checkpoint-113000/rng_state_13.pth +3 -0
  12. checkpoint-113000/rng_state_14.pth +3 -0
  13. checkpoint-113000/rng_state_15.pth +3 -0
  14. checkpoint-113000/rng_state_2.pth +3 -0
  15. checkpoint-113000/rng_state_3.pth +3 -0
  16. checkpoint-113000/rng_state_4.pth +3 -0
  17. checkpoint-113000/rng_state_5.pth +3 -0
  18. checkpoint-113000/rng_state_6.pth +3 -0
  19. checkpoint-113000/rng_state_7.pth +3 -0
  20. checkpoint-113000/rng_state_8.pth +3 -0
  21. checkpoint-113000/rng_state_9.pth +3 -0
  22. checkpoint-113000/scheduler.pt +3 -0
  23. checkpoint-113000/trainer_state.json +3 -0
  24. checkpoint-113000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -53,3 +53,4 @@ checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
53
  checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
53
  checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
+ checkpoint-113000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-113000/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": false,
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-113000/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-113000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:feaf1a5571e4ae74080d871c535affab470d0d2857d7a94a6e1dfdbfe620fa2c
3
+ size 4538452374
checkpoint-113000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71dd7df87dbe1d9ea9c6d1a23440c489d3546aca9440cb78568d24857cc61ab2
3
+ size 6630551906
checkpoint-113000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9aabc21b275a6c3e676310481e5d7f0784534e1fa8b9927191c9b720f6482ef7
3
+ size 15984
checkpoint-113000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7833eb5cb8b12c1d39d1935613b1486b9344cc493cd13dab85a45f8c1eb5707
3
+ size 15984
checkpoint-113000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88274b5933f9abe98071fce8bfddb736de9899a6b1e0990a35cfba9c46e602fe
3
+ size 15997
checkpoint-113000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9156f31bd2e3be5290af5163b7426d1a8f0fa7abe9e0c51a866d30aafafe34d0
3
+ size 15997
checkpoint-113000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e044b77a72743b788cd26fa9088e4ec41cdf12799d3fbb4ba1fb9b4d4603cded
3
+ size 15997
checkpoint-113000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:751b4196d96dbf0647571ebc13665aaf949e0b429baa891111e66e253e6f9533
3
+ size 15997
checkpoint-113000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79d98302f9c01de11be4c0520693ab8a398e1b0926e511e3f362b0ff040087ea
3
+ size 15997
checkpoint-113000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d3e3238252e43d510766d2c2f569bdf5b49e1ee9d012b5dbbb816cde07210c4
3
+ size 15997
checkpoint-113000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22eac456cabb1f906d2583999bc16d1fb317f4a09dff2d0569c2936675573cec
3
+ size 15984
checkpoint-113000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a53ad73431ea6dea020edb996c66703d93c3fd624f4ecc7d2b4bd86553e799c0
3
+ size 15984
checkpoint-113000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:242212a3cae0bfafdb1266ae373088860eb44f99ca3b822588c02b33291e081f
3
+ size 15984
checkpoint-113000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0a67f2b60bca04f86e1abbe420dedb7b88ee799330fda8c0290245d93f84016
3
+ size 15984
checkpoint-113000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d880cc847e8fa2d7fc0edbb4ad7e95f74ea3eb0d0c1c1994f951597acd5115ba
3
+ size 15984
checkpoint-113000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6a1b221168e1c33e53e160580ca301e36c73eddfa1ef4a8a2135c894575e89b
3
+ size 15984
checkpoint-113000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:053dc3c93185e34602698e0ce49e5cd465cfef064d44ad407d1c1909096a7e81
3
+ size 15984
checkpoint-113000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80c6282c61683a9ae8fe4bced781d82a2d68322755acb6cc07c4dd8977599272
3
+ size 15984
checkpoint-113000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41e5ad562e42a2b52d56b1d44476f8738ff14738d2106bbcf539477afc484db8
3
+ size 1064
checkpoint-113000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5998804372cafeaa9ed03cf19f5a8c8c3f8f78db7db5add897f5c953bedb8d15
3
+ size 17489084
checkpoint-113000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f813ea370cedfa4e2438e1e5f9515666409584e1aa8fea65e11571de1056736
3
+ size 5944