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
60951b9
·
verified ·
1 Parent(s): 0a6ae9e

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-145000/config.json +79 -0
  3. checkpoint-145000/generation_config.json +6 -0
  4. checkpoint-145000/model.safetensors +3 -0
  5. checkpoint-145000/optimizer.pt +3 -0
  6. checkpoint-145000/rng_state_0.pth +3 -0
  7. checkpoint-145000/rng_state_1.pth +3 -0
  8. checkpoint-145000/rng_state_10.pth +3 -0
  9. checkpoint-145000/rng_state_11.pth +3 -0
  10. checkpoint-145000/rng_state_12.pth +3 -0
  11. checkpoint-145000/rng_state_13.pth +3 -0
  12. checkpoint-145000/rng_state_14.pth +3 -0
  13. checkpoint-145000/rng_state_15.pth +3 -0
  14. checkpoint-145000/rng_state_2.pth +3 -0
  15. checkpoint-145000/rng_state_3.pth +3 -0
  16. checkpoint-145000/rng_state_4.pth +3 -0
  17. checkpoint-145000/rng_state_5.pth +3 -0
  18. checkpoint-145000/rng_state_6.pth +3 -0
  19. checkpoint-145000/rng_state_7.pth +3 -0
  20. checkpoint-145000/rng_state_8.pth +3 -0
  21. checkpoint-145000/rng_state_9.pth +3 -0
  22. checkpoint-145000/scheduler.pt +3 -0
  23. checkpoint-145000/trainer_state.json +3 -0
  24. checkpoint-145000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -68,3 +68,4 @@ checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
68
  checkpoint-142000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
69
  checkpoint-143000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
70
  checkpoint-144000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
68
  checkpoint-142000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
69
  checkpoint-143000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
70
  checkpoint-144000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
71
+ checkpoint-145000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-145000/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-145000/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-145000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a6fe1040dc2da2e51904b047b0b3bb4a07c0f8f0e6b6fe415a4875531a90c50
3
+ size 4539487822
checkpoint-145000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a9639472252f4f6ed50f395989b12b2f5a7bab7a201c4e4f21c1ab9c0f882fb
3
+ size 6632345482
checkpoint-145000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c6b966bcb66dfaa902ae460da8a7a87c6184ecf7114dfb9e9da66f10337a46d
3
+ size 15984
checkpoint-145000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6acd40099e8c92da4dac65c9c367a5f5d9f3813eb771c820ac876b670c153c4a
3
+ size 15984
checkpoint-145000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2109a50fd2939b7b58a29c7c46d54e2a40e132aaf4195a58afcd0f8852a0fcc5
3
+ size 15997
checkpoint-145000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6daed0c6301833b501d3582afff1af337ced14208bb4b4b742c45f8d67248eee
3
+ size 15997
checkpoint-145000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b17c0ce1dd003e0d01f2434a30ed377d20cf5baa143c81668f4ee4f25675e5fe
3
+ size 15997
checkpoint-145000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75d133b16a3438ca6db44e0d39759388fb81839d3c0dce752e0fb4f71de83a23
3
+ size 15997
checkpoint-145000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:084386fcd7bfae15f6cd8bcf08a8a27b864a61b66fe71c8c60e5a4785415a2d8
3
+ size 15997
checkpoint-145000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae285ed2dc72c6a842d6c45188489ac2a802f6b82dfae280a7dde946245589fc
3
+ size 15997
checkpoint-145000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6af950422309ff6fc0f518745f3c085cad636e510cb167588a5b8c1dc8a6a2a
3
+ size 15984
checkpoint-145000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81d656af7da32be7caf1601a468a8e73cc48f2b310744b266c74c4afdb7fc3a6
3
+ size 15984
checkpoint-145000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cac800bf80bea0a7a752604639d1ce9a9df8f4296df6c41ce494dcfa80f7307a
3
+ size 15984
checkpoint-145000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a729abe1097924ca7ac1d62675b29983f1a114c1b024a78e3671c2ba460c3ab9
3
+ size 15984
checkpoint-145000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d93e7800206833e7f6d427155c30bd4740eda06efbe4f74fbe356ce1e773193
3
+ size 15984
checkpoint-145000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3ff08d8a987ea0fa9974c24ef7e8d451ce6b9db3e43f1fc4dc55a2c2bce6064
3
+ size 15984
checkpoint-145000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a61a42d56810f77c752ded5450f33217e634cc458606ee13eb650a63869453d
3
+ size 15984
checkpoint-145000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:703dbe08b90105eaf4da772c7d154bc210bc02160600354979aae2283ec33275
3
+ size 15984
checkpoint-145000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1afdcb2c3418b433922666d3aadbedcd40d7eef9efeebfbba224fd08f25b7ce8
3
+ size 1064
checkpoint-145000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52b946ba5b6dc19c629a675f3f9b21c3c3ef37168e6931ee74a41d49b4ff6fb3
3
+ size 22415037
checkpoint-145000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71d13cef18bc4d8946264d52f7383e8d1481ce56f5b5d0587db6c20561757c90
3
+ size 6008