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
9b78ec0
·
verified ·
1 Parent(s): 9102c71

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-170000/config.json +79 -0
  3. checkpoint-170000/generation_config.json +6 -0
  4. checkpoint-170000/model.safetensors +3 -0
  5. checkpoint-170000/optimizer.pt +3 -0
  6. checkpoint-170000/rng_state_0.pth +3 -0
  7. checkpoint-170000/rng_state_1.pth +3 -0
  8. checkpoint-170000/rng_state_10.pth +3 -0
  9. checkpoint-170000/rng_state_11.pth +3 -0
  10. checkpoint-170000/rng_state_12.pth +3 -0
  11. checkpoint-170000/rng_state_13.pth +3 -0
  12. checkpoint-170000/rng_state_14.pth +3 -0
  13. checkpoint-170000/rng_state_15.pth +3 -0
  14. checkpoint-170000/rng_state_2.pth +3 -0
  15. checkpoint-170000/rng_state_3.pth +3 -0
  16. checkpoint-170000/rng_state_4.pth +3 -0
  17. checkpoint-170000/rng_state_5.pth +3 -0
  18. checkpoint-170000/rng_state_6.pth +3 -0
  19. checkpoint-170000/rng_state_7.pth +3 -0
  20. checkpoint-170000/rng_state_8.pth +3 -0
  21. checkpoint-170000/rng_state_9.pth +3 -0
  22. checkpoint-170000/scheduler.pt +3 -0
  23. checkpoint-170000/trainer_state.json +3 -0
  24. checkpoint-170000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -78,3 +78,4 @@ checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
  checkpoint-161000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
  checkpoint-168000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
78
  checkpoint-161000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
  checkpoint-168000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
+ checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-170000/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-170000/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-170000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32c29dae7fbd08cd8df789c73a00f36906bbfa48defa9351a7db7521147e03ac
3
+ size 4539487822
checkpoint-170000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83fa4ca69222f75d3101d78e5b17df7dbf3aabc30b3dd7646d622656a9e5dddb
3
+ size 6632345482
checkpoint-170000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98d1e1b7b0f146465ac9fab93df11a34bad3812526b81204121b39ea733111fa
3
+ size 15984
checkpoint-170000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:528244a7e5c4c9164bd4ac57d26e4a3cf4a53c63bfb409b82aefe3750948183b
3
+ size 15984
checkpoint-170000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54e4805753b963060785c60e7755154748958cc7cb018b738bce157f5c0bf6ba
3
+ size 15997
checkpoint-170000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2c08bf8db8e642f3987085a9d16821f740035c8c3d2ffb6b79d7b49c1bdd119
3
+ size 15997
checkpoint-170000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:968654a123b13c5660cd8b24b0feca9f3c1580a2b9532323463d92c0c3cff3dd
3
+ size 15997
checkpoint-170000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:003a79d38106b4bc5afa33abae70bc0e923662aa29aaf1b445222b9a28b1e5c4
3
+ size 15997
checkpoint-170000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72659e2b82deee814b5aac73ec258b546ff6792b3015851f8e50f9beeb254334
3
+ size 15997
checkpoint-170000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d09eb1057d5c165c0cdf7f20ae70f0a4eed58d2fbfbffd13b23f7c275f15d5a3
3
+ size 15997
checkpoint-170000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d281b59b82c694198244e06250c30a57f67064495201b1f7f31d605eccdec4c5
3
+ size 15984
checkpoint-170000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb080dc7d0ef30b045ca980ac23c5c3c6336e1758262c01ec77bcb61602bcef7
3
+ size 15984
checkpoint-170000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0178f1a7572d166072f3830feaf8f0bf9c5e108283ce7d23391534aa104bd684
3
+ size 15984
checkpoint-170000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4ebc7a55f0b97ef1c77bbd51968fc5b0c274c39a2b6755285d0abb9e7da121a
3
+ size 15984
checkpoint-170000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc94dd2be13435410dffb00cb4ce83b743126b5933c53f5d1f56fff3d29203db
3
+ size 15984
checkpoint-170000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:563dccd773eda5849ca6574897735f1b7423f5cba7962c7d7faf1fc4099a8108
3
+ size 15984
checkpoint-170000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0827cdb12d7378f4da1df285360b6685082a03dbe02cb446a3ef09c7c64aa12
3
+ size 15984
checkpoint-170000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2cb2387528367c511a7d0214e9ee1fc3ed2a9748c58649ed932b3e84c0af7f2b
3
+ size 15984
checkpoint-170000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:230cfffa87b85649ed86bd4b446d61bf13735f62ba43398a798cf217a49f21e8
3
+ size 1064
checkpoint-170000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:326a5281a39b379695bcdbe955c3e6a23716250315b83efe768b55adbfcf7884
3
+ size 26290615
checkpoint-170000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4df5a2e22846b993e7a1ab59db8f1e9870011f130aa8b633e164c64d3829edf2
3
+ size 6008