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
2797c86
·
verified ·
1 Parent(s): b93b05b

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-40000/config.json +79 -0
  2. checkpoint-40000/generation_config.json +6 -0
  3. checkpoint-40000/model.safetensors +3 -0
  4. checkpoint-40000/optimizer.pt +3 -0
  5. checkpoint-40000/rng_state_0.pth +3 -0
  6. checkpoint-40000/rng_state_1.pth +3 -0
  7. checkpoint-40000/rng_state_10.pth +3 -0
  8. checkpoint-40000/rng_state_11.pth +3 -0
  9. checkpoint-40000/rng_state_12.pth +3 -0
  10. checkpoint-40000/rng_state_13.pth +3 -0
  11. checkpoint-40000/rng_state_14.pth +3 -0
  12. checkpoint-40000/rng_state_15.pth +3 -0
  13. checkpoint-40000/rng_state_2.pth +3 -0
  14. checkpoint-40000/rng_state_3.pth +3 -0
  15. checkpoint-40000/rng_state_4.pth +3 -0
  16. checkpoint-40000/rng_state_5.pth +3 -0
  17. checkpoint-40000/rng_state_6.pth +3 -0
  18. checkpoint-40000/rng_state_7.pth +3 -0
  19. checkpoint-40000/rng_state_8.pth +3 -0
  20. checkpoint-40000/rng_state_9.pth +3 -0
  21. checkpoint-40000/scheduler.pt +3 -0
  22. checkpoint-40000/trainer_state.json +0 -0
  23. checkpoint-40000/training_args.bin +3 -0
checkpoint-40000/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-40000/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-40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48e3356756ee6e0d2d8d8b721911d577ec3f7733101947106f9965fddebabeef
3
+ size 4539487822
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c11c79aff674de31161a9673290f894f86263d392682434e16806d20904ad47
3
+ size 6632345482
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d22fbf97aa93eb8cf41b91848e9cc7f89d5516ad2dc97b2417d938180cfce0d
3
+ size 15984
checkpoint-40000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2dceeaa07b8d1d0078052f3236b50a1ed0b0e0ad88c5b9934b82158d3dc5714
3
+ size 15984
checkpoint-40000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffa3feabe817131442ebd3b35ef29f9aab6d91f2186fb4960fd4cd7e025eee29
3
+ size 15997
checkpoint-40000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42d826898fed29dc0d372425aead71755ae9d175815928e211708278baea4289
3
+ size 15997
checkpoint-40000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72f7dbe5daed92619f4c256d1059eda2c5b346994dc49359574c371b1f42e2cd
3
+ size 15997
checkpoint-40000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7453ce462d1139eae066648c1c65d953b04184cec9bf01c6b474ddbab0655478
3
+ size 15997
checkpoint-40000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eff79123fc0ca2bb1fb1f5dfcf0d17ae51e47dbce20c21a8d405f3ad352c2865
3
+ size 15997
checkpoint-40000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ca1246282c3bd8370d35f9d9a95b6e5880668ce4a482ddf237cd5408a4f13a4
3
+ size 15997
checkpoint-40000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18e29fd66e76c61c6a1c01e1668342b7158bdd3f58d66f76edceb8852a12b2fa
3
+ size 15984
checkpoint-40000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82dd6e50880974a94e705b8619c742aef81aa4bb9dd7182a370bd7c086b31280
3
+ size 15984
checkpoint-40000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1833ad119d371379cedb0fbc746888ddc0587df4ef2d6c3631937ce753d63b61
3
+ size 15984
checkpoint-40000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b83cf257c22af4070230c4bf17a63c8b5502a6da29e07c500ea2d74910e31673
3
+ size 15984
checkpoint-40000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f78173c0cccb5a541faf821cf0e5f402bb6103d40692decd50c10585e4146c3
3
+ size 15984
checkpoint-40000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54a4ef04d8c89765274f4397ef626e1c0a9b1fdc2a5ca0b5f1d6ab4c21ab4382
3
+ size 15984
checkpoint-40000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd037cf42b022b0a0887db31d241bf92121316f782335374134b6cee3fb53169
3
+ size 15984
checkpoint-40000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81f22dc8add53c2fc6b2244e777d20d59ce3876bfeb8a327943574db31663a5c
3
+ size 15984
checkpoint-40000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:825e1818d8a0f8348d506acdab26ab27594e76e01808880216f9807387891534
3
+ size 1064
checkpoint-40000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-40000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71e91e1f0c2c00de1919084a09d0c4973390e647d4801a7f51cdf9b750836095
3
+ size 6008