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
2619cf2
·
verified ·
1 Parent(s): e1e8d10

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-65000/config.json +79 -0
  2. checkpoint-65000/generation_config.json +6 -0
  3. checkpoint-65000/model.safetensors +3 -0
  4. checkpoint-65000/optimizer.pt +3 -0
  5. checkpoint-65000/rng_state_0.pth +3 -0
  6. checkpoint-65000/rng_state_1.pth +3 -0
  7. checkpoint-65000/rng_state_10.pth +3 -0
  8. checkpoint-65000/rng_state_11.pth +3 -0
  9. checkpoint-65000/rng_state_12.pth +3 -0
  10. checkpoint-65000/rng_state_13.pth +3 -0
  11. checkpoint-65000/rng_state_14.pth +3 -0
  12. checkpoint-65000/rng_state_15.pth +3 -0
  13. checkpoint-65000/rng_state_2.pth +3 -0
  14. checkpoint-65000/rng_state_3.pth +3 -0
  15. checkpoint-65000/rng_state_4.pth +3 -0
  16. checkpoint-65000/rng_state_5.pth +3 -0
  17. checkpoint-65000/rng_state_6.pth +3 -0
  18. checkpoint-65000/rng_state_7.pth +3 -0
  19. checkpoint-65000/rng_state_8.pth +3 -0
  20. checkpoint-65000/rng_state_9.pth +3 -0
  21. checkpoint-65000/scheduler.pt +3 -0
  22. checkpoint-65000/trainer_state.json +0 -0
  23. checkpoint-65000/training_args.bin +3 -0
checkpoint-65000/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-65000/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-65000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39b2e63dcaaf70d93afe5a2974a4adb68eed597adf3a757daec65c9a5695170e
3
+ size 4539487822
checkpoint-65000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d647af0823b65f476a19e0838626c93850af86667b01ea0713a3a6b6a49b4f32
3
+ size 6632345482
checkpoint-65000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a423c43f82971d52dd7172a0cfbba56ab1a247fbadb5c5ac25988e0a9e8bc0cc
3
+ size 15984
checkpoint-65000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12866d42513b03e73619dd7b399a8b98f3d8ec78692d16c3564df3ba5ae505d5
3
+ size 15984
checkpoint-65000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f244046e7d89bc13f0bc9eca90cab7231b069d51fc0cdd2eff3e646bfd431567
3
+ size 15997
checkpoint-65000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1f7029aecc6d79ee8ba4544dd7978ed81af55dea0ed177bbef5a6678d289806
3
+ size 15997
checkpoint-65000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6bbba526b7af78c516c304d2b9d0df4b009f4e28f8a1cb5f503f33225d83931
3
+ size 15997
checkpoint-65000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19930c17eb680d98b45f0cfb005079dc63ee27480f6da514eababf46c0140062
3
+ size 15997
checkpoint-65000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a8cad964c3a72d55b7213ed63b10230cb8c69752658ec02ebc65dafd1912100
3
+ size 15997
checkpoint-65000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d2b1869e2c5806982cc805e0b1619ea1a0c1ffb6501b51cb6906ccf64030adc
3
+ size 15997
checkpoint-65000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e27be88953fb147c560acd43d19f7565bfb6780bf95436c9c9e53accb5666b5
3
+ size 15984
checkpoint-65000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:081bfa61b36052c9dd9a0bc771d6155fab73e78f28a464e805a19edb618c5986
3
+ size 15984
checkpoint-65000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9aeeef02689e0ba102c0de8503bacc91f161dcf76f62e83fa4a0d50c8308bb
3
+ size 15984
checkpoint-65000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f0aee2ad730efb4eb93407dec29fe596f0f5ed1d200d137caf2e401b3204ec2
3
+ size 15984
checkpoint-65000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf50dd85fe224b4a019dc983374682358a364d75e1fc1c90ad59f1b3493a0304
3
+ size 15984
checkpoint-65000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64c73daa6155c7e72cda4599b272e531732418e9b25294cc8473cbc38ce81452
3
+ size 15984
checkpoint-65000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec1200354c0ce4ecc133e2a9a40d760507fabbf6eb730f8b455c146934fd6175
3
+ size 15984
checkpoint-65000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44a486d5e3c653b4aefd2a8bdaaa40ba99b4f5e72735ac7af0267d71e908ce60
3
+ size 15984
checkpoint-65000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32c8bd471103c5c5e584c02e5a955003abc1664f11d5133e48d5e15e24264c86
3
+ size 1064
checkpoint-65000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-65000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61c3a70de3dcb1ed17e6a32f9f6c066cc0d83b6be6fc19e2ade4b22e85ce283f
3
+ size 6008