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
aaf20f1
·
verified ·
1 Parent(s): bb79198

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-90000/config.json +79 -0
  3. checkpoint-90000/generation_config.json +6 -0
  4. checkpoint-90000/model.safetensors +3 -0
  5. checkpoint-90000/optimizer.pt +3 -0
  6. checkpoint-90000/rng_state_0.pth +3 -0
  7. checkpoint-90000/rng_state_1.pth +3 -0
  8. checkpoint-90000/rng_state_10.pth +3 -0
  9. checkpoint-90000/rng_state_11.pth +3 -0
  10. checkpoint-90000/rng_state_12.pth +3 -0
  11. checkpoint-90000/rng_state_13.pth +3 -0
  12. checkpoint-90000/rng_state_14.pth +3 -0
  13. checkpoint-90000/rng_state_15.pth +3 -0
  14. checkpoint-90000/rng_state_2.pth +3 -0
  15. checkpoint-90000/rng_state_3.pth +3 -0
  16. checkpoint-90000/rng_state_4.pth +3 -0
  17. checkpoint-90000/rng_state_5.pth +3 -0
  18. checkpoint-90000/rng_state_6.pth +3 -0
  19. checkpoint-90000/rng_state_7.pth +3 -0
  20. checkpoint-90000/rng_state_8.pth +3 -0
  21. checkpoint-90000/rng_state_9.pth +3 -0
  22. checkpoint-90000/scheduler.pt +3 -0
  23. checkpoint-90000/trainer_state.json +3 -0
  24. checkpoint-90000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -42,3 +42,4 @@ checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-81000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  checkpoint-89000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
42
  checkpoint-81000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  checkpoint-89000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
45
+ checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-90000/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-90000/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-90000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41d57dd352c55192375c3af08700cc6641b8ca5531cb40151dd2442e4048ca0e
3
+ size 4539487822
checkpoint-90000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2c852418a2420412d3fdc720102f7e7a048ceac6129ebbbf82b588b7bcca367
3
+ size 6632345482
checkpoint-90000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f73d4806f8cbf99b1adebb6475e304d5c1744b81b33c4702881924928623ce8
3
+ size 15984
checkpoint-90000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea80acf94413290d3b68a9959d05dbe3b91535115b6f20ac7fb78b9cdd9b709f
3
+ size 15984
checkpoint-90000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a1b8f0e7a5602ed4fffe9407d3b1f353d1a3e1bb96c85ef70c69f22f98e3440
3
+ size 15997
checkpoint-90000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:445ac985203a8d9fbc6349906ff51602a37249a00ed724aeee88df37ca84c3c6
3
+ size 15997
checkpoint-90000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc4f22881e3028e988b9e9bd100794aa288f030242828091c75336e5b6f7008a
3
+ size 15997
checkpoint-90000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d6195f4f96f80c1d5bd8b8e3854f277152365416dcd9bbaef9dd79c7053488f
3
+ size 15997
checkpoint-90000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b05be6e99fa232f09386354db01ec040699f4a97ce66f1d3b638eb8a472d02fd
3
+ size 15997
checkpoint-90000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59e8f48a52c5d8df169f3f3fb9543b0a419ac58a78efce6e866781355f8439bf
3
+ size 15997
checkpoint-90000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9613892197542e0a6b662cd7a64a3acaafa63eda54055c149475bec26a4f519b
3
+ size 15984
checkpoint-90000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b9aaf41885a47eb7e40c7cbb3d458ec2e4003358ed788a7aa79963389227828
3
+ size 15984
checkpoint-90000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ac99079968ad6a89f91df0733145205815fe3ab608239257af9c9b31b5572fb
3
+ size 15984
checkpoint-90000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3979ca24526a944dd6bd5106f059e41a8104bb4d7be4cb1db4ce78a9f910597c
3
+ size 15984
checkpoint-90000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f10740672878cb29522fc12241662e7d225b6cf8b0829a49dd74e3eded5cb793
3
+ size 15984
checkpoint-90000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:637907f733de127620193476d766edb185ea059bf1b21512a3fcdf6d1c37884f
3
+ size 15984
checkpoint-90000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53dc0c5578dae52fd025aee79b5eeef0827d70c3a170d1eed2ee6fbdc0bd84a0
3
+ size 15984
checkpoint-90000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c037177e71dae597e40594cbc2747ab8fceef640a9984141010e8f933bfb251a
3
+ size 15984
checkpoint-90000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f17fac99609a9eef75b7ab41b9b39e47bf03a67a9c70b8a6c43cc3eb54f185ac
3
+ size 1064
checkpoint-90000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:824ae3d7fd35d9af0e920404dfafd0467ed667d3bffbabc995bafb0d336b35dd
3
+ size 13885470
checkpoint-90000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3a71dd25996700a2cabc2a7712e9edd0c768e9fdee21d2df677e9faaecceda6
3
+ size 6008