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
010e81e
·
verified ·
1 Parent(s): 589c8e2

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-101000/config.json +79 -0
  3. checkpoint-101000/generation_config.json +6 -0
  4. checkpoint-101000/model.safetensors +3 -0
  5. checkpoint-101000/optimizer.pt +3 -0
  6. checkpoint-101000/rng_state_0.pth +3 -0
  7. checkpoint-101000/rng_state_1.pth +3 -0
  8. checkpoint-101000/rng_state_10.pth +3 -0
  9. checkpoint-101000/rng_state_11.pth +3 -0
  10. checkpoint-101000/rng_state_12.pth +3 -0
  11. checkpoint-101000/rng_state_13.pth +3 -0
  12. checkpoint-101000/rng_state_14.pth +3 -0
  13. checkpoint-101000/rng_state_15.pth +3 -0
  14. checkpoint-101000/rng_state_2.pth +3 -0
  15. checkpoint-101000/rng_state_3.pth +3 -0
  16. checkpoint-101000/rng_state_4.pth +3 -0
  17. checkpoint-101000/rng_state_5.pth +3 -0
  18. checkpoint-101000/rng_state_6.pth +3 -0
  19. checkpoint-101000/rng_state_7.pth +3 -0
  20. checkpoint-101000/rng_state_8.pth +3 -0
  21. checkpoint-101000/rng_state_9.pth +3 -0
  22. checkpoint-101000/scheduler.pt +3 -0
  23. checkpoint-101000/trainer_state.json +3 -0
  24. checkpoint-101000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -47,3 +47,4 @@ checkpoint-96000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
47
  checkpoint-97000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
48
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
49
  checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
47
  checkpoint-97000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
48
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
49
  checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
50
+ checkpoint-101000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-101000/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-101000/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-101000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e71b8ec71c24ed278ba346459c7cb84a3a52065cf3cbc61381c61bca076234e2
3
+ size 4538452374
checkpoint-101000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e08a3fd3fb59ae64eeef802dff3ae6257616502e9661a7f273157eafcef1f66
3
+ size 6630551906
checkpoint-101000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcc36c9a169def50f11ab2be2d9365e0527c920a1f22906905fdb73e541dbb1b
3
+ size 15984
checkpoint-101000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2cf3bad930c9d654954302ae0f66593432e1e07f1ff4126b0ded5abc1bfb10df
3
+ size 15984
checkpoint-101000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:962d62c24606bfa653e90fae487b588ec44ffc70012e368aabf227ce3eb8511d
3
+ size 15997
checkpoint-101000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a84bf01354b5f3028d2cd3a9e739760092fabf70fadbbd86ff6ca4af116a8b47
3
+ size 15997
checkpoint-101000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f785c9a248b7f8d08dce0033456d345ad72586172dcb2c080f9922d73ddc87de
3
+ size 15997
checkpoint-101000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe61a6991edbaad9fbc4b317675304b7df68037206d19a075ef70bfa01eb1136
3
+ size 15997
checkpoint-101000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:054ad1cf12c4cbf28c84d4b019417d85ba870720fcaf8e3065c1a661f54d7e0f
3
+ size 15997
checkpoint-101000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81c07753ccc9fb9de4766acf5f245e562ad755a0f10113102520a07cd6cf93e4
3
+ size 15997
checkpoint-101000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbb940c3ed81fae5c646e872e654641b9f8a9251aca348886b7ff0139b7dad4a
3
+ size 15984
checkpoint-101000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b574c8c4cb5229e11c393bb36d01803185e463ce5969fd11069e253cb5263422
3
+ size 15984
checkpoint-101000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69ffbb3e1318ed01f252db4c5ae064af1465ef96df26be61ac426656aeca157d
3
+ size 15984
checkpoint-101000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68c3870803a68926fd0658645372d09dc90b7107e0bff1738126c5e610ecda1d
3
+ size 15984
checkpoint-101000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3af149b818442f54d41843091456c662cf26871fc8b396c9bb03ac1ad306f73f
3
+ size 15984
checkpoint-101000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b37513d088bdadf124b581f81915a1fde302fbe50a1d440c6eb007778f293b2
3
+ size 15984
checkpoint-101000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea2969514ecfd90c19d09ba06458bb68f00a8567f95cae04dc02b67b2fcbd4c7
3
+ size 15984
checkpoint-101000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:531e58606b9100e84cdec53e501359ffec26c4b5ebda2ebad55b8df069ef06e0
3
+ size 15984
checkpoint-101000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebaee7f7d16310fcd3305bdb81dc32d8e319f4ca73404af3984b2d0ea5569274
3
+ size 1064
checkpoint-101000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87fa06716a4d711b134afc35608ba1650c0cd097638097d11c38530630101179
3
+ size 15621008
checkpoint-101000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3321f5404a8c15de296a04b82541723bbf8bfaff5479f5be3e6b8f327b5ce5c4
3
+ size 5944