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
2d08f0e
·
verified ·
1 Parent(s): 4ba8d3b

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": 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-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 sha256:54035838fa70c3fbc5e5a3f606a8750a75f71c20fb296d35c177e60865cc8d05
3
+ size 4539487822
checkpoint-101000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a26595d5a16ff18d2196ea7879162d00996f38699582dff6b78c5db639738c1b
3
+ size 6632345482
checkpoint-101000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f8f9da11c6f53799096d8408de0ea66ddee828d6d38dd40026b4fc814c63c66
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 sha256:9df3b8161f72ccd1667b047c36d882bab51b245e7843d3366a6f38a0eb948203
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:dff51d7532e2eabc07c9173102bf389a3a150624cc37ff7ccea90ef63f4191db
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 sha256:0745a2a178ad5f658a2f3b13cbcff7cc365f6118552f18a2c3b4334291091f4c
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 sha256:3dadced8b1c040f8f8c1eeae58171824dba2e1887afe5fcbdc47620be69b1742
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:efecef4557b58e7757f1598a258b4075d688740bdd760acee5040e30c45940d2
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:c9314704a7674faf8d6546b511cba142404500b1e2101b6925ccc32f7902c7da
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:a931ebe1948bfd380654b12adf117615f5b34016f022e67933fd84dfa803433e
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 sha256:3fba3ed0575ba570d00c903b8c3e4bd2e01b2349545d57e209eda3e117fa7e13
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:f680fd702b3a7f75f09f6b8ea9d432cd7d277099028c61ddeadb4162bc7e4d77
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:a4a55c50267593505dbb472d47b2360d33be54525b7240b115d96b6cfe15cedb
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 sha256:9357e93505cd20ad7cbf9de148d936d1444f42797b81dd01459673bf897de2e1
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 sha256:19af3dd7e5e2d937fb21df295a4558ebdb4b949d8d2366cc0b2e8250db6af4ce
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 sha256:0567caaef50a3a0abf6c5976b8b6c84e722ed729ea56bd04cf5f22ef560ade4e
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 sha256:0fbe440d40cecc77f8ed427f3f03538f60225bb33da306d6d5f9d1acaf267566
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 sha256:9af37879ebc6c3b0754602c921b239434dfb53f8f44f6b1d5fe5a911f91fe949
3
+ size 15984
checkpoint-101000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4eeb353488b1fb037425166bfbd76871e6c1acd6ff7fe736b0fed11167450b0f
3
+ size 1064
checkpoint-101000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:513442ee3e81c9806af9f2647c37a6d98b93cbdf0d63dda59212d4254c60166b
3
+ size 15582922
checkpoint-101000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd87377984a02cccf3038a050012e474eb160a436e20bdbfc037d50b8ffb972a
3
+ size 6008