Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_aug_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
9371b5d
·
verified ·
1 Parent(s): 5f7a8c9

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-102000/config.json +79 -0
  3. checkpoint-102000/generation_config.json +6 -0
  4. checkpoint-102000/model.safetensors +3 -0
  5. checkpoint-102000/optimizer.pt +3 -0
  6. checkpoint-102000/rng_state_0.pth +3 -0
  7. checkpoint-102000/rng_state_1.pth +3 -0
  8. checkpoint-102000/rng_state_10.pth +3 -0
  9. checkpoint-102000/rng_state_11.pth +3 -0
  10. checkpoint-102000/rng_state_12.pth +3 -0
  11. checkpoint-102000/rng_state_13.pth +3 -0
  12. checkpoint-102000/rng_state_14.pth +3 -0
  13. checkpoint-102000/rng_state_15.pth +3 -0
  14. checkpoint-102000/rng_state_2.pth +3 -0
  15. checkpoint-102000/rng_state_3.pth +3 -0
  16. checkpoint-102000/rng_state_4.pth +3 -0
  17. checkpoint-102000/rng_state_5.pth +3 -0
  18. checkpoint-102000/rng_state_6.pth +3 -0
  19. checkpoint-102000/rng_state_7.pth +3 -0
  20. checkpoint-102000/rng_state_8.pth +3 -0
  21. checkpoint-102000/rng_state_9.pth +3 -0
  22. checkpoint-102000/scheduler.pt +3 -0
  23. checkpoint-102000/trainer_state.json +3 -0
  24. checkpoint-102000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -45,3 +45,4 @@ checkpoint-92000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
45
  checkpoint-94000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
46
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
47
  checkpoint-98000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
45
  checkpoint-94000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
46
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
47
  checkpoint-98000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
48
+ checkpoint-102000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-102000/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-102000/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-102000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd14b442092f337bbd681e0f46d9dfb464aa0465126edd978484d6ab7f6c231e
3
+ size 4539487822
checkpoint-102000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6a95eda8f56690d6c4a56a5c2269bd139b24919a66db5caf66a8721b4d25b0f
3
+ size 6632345482
checkpoint-102000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22a3d252a4922162f95c10a5c314f148fed15bd90e24450c427fa767fbf2090e
3
+ size 15984
checkpoint-102000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79366371c34fbb54c902abcf64b44bf09a534101bec7a633684d7153b7132ac6
3
+ size 15984
checkpoint-102000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1711fa80b52c0c69937b7e8330f0acec69896c6d1afa201843aaca5324699201
3
+ size 15997
checkpoint-102000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e384340948f234db81ab4c1339b37490bfdb94deaca1947a951ad02d6e8f6a3
3
+ size 15997
checkpoint-102000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:784c422704246cdaa99878563484511148fc4d1c54f455a3ac4684030c68dca0
3
+ size 15997
checkpoint-102000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79c0dd5ca02f1c1c5ab6082ee7c4ec2115f0e8ee21a1171a881426b87f1794ad
3
+ size 15997
checkpoint-102000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c1c47d22298dbfb7706bab958b9fe35913105e3453547c2c677b75609d6490b
3
+ size 15997
checkpoint-102000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d33a309dd35d5adf5621e6a10b00f1b3cdc005de360378e4793842dde79ce1c5
3
+ size 15997
checkpoint-102000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:891ff6cba873c38b3eff255299e6fa59b115d5afdf568dc09bd908cb97aaf237
3
+ size 15984
checkpoint-102000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca6b8c19d62ce5148f277d91f90acada58b7bef650f898f01a941d9e3dea5ec4
3
+ size 15984
checkpoint-102000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8efc067b804e794ffec8d3d15a7b8e9e2fb28fe6c439f71ed76b642427b09c5
3
+ size 15984
checkpoint-102000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b19e5808b10aa77201ae3410d9fb9d433b36c07526d2dd2a76458df44003451b
3
+ size 15984
checkpoint-102000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12148627fdb4c83ddf52f065c61a76e734d35e7843862ad9b26ad233429304d8
3
+ size 15984
checkpoint-102000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:860f19e2be69e50598213d758b50df28444cfbc5ff02f8c30b552b904831b67d
3
+ size 15984
checkpoint-102000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:798934adb5a5eb8c98019629322b85748af1343e43be98e8f7943ea7c1de57cb
3
+ size 15984
checkpoint-102000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea17469eaadfff41f0348dff1f89af3e9e2dafa74253e9ff76af66f1ef16792d
3
+ size 15984
checkpoint-102000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88c6cf913ca279c15057bf36dbeb75a89aeb132dbf47442b7435ffaba7772f8e
3
+ size 1064
checkpoint-102000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97b681960ee88df9da8939e6ae6c199016f52dd637a5afcae3ee42035523cb24
3
+ size 15733508
checkpoint-102000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62df80d1346b0241886bd644d859e6384c7dd8643e538b3c01432ede852296e3
3
+ size 6008