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
cf0da0f
·
verified ·
1 Parent(s): 8bdddc7

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-194000/config.json +79 -0
  3. checkpoint-194000/generation_config.json +6 -0
  4. checkpoint-194000/model.safetensors +3 -0
  5. checkpoint-194000/optimizer.pt +3 -0
  6. checkpoint-194000/rng_state_0.pth +3 -0
  7. checkpoint-194000/rng_state_1.pth +3 -0
  8. checkpoint-194000/rng_state_10.pth +3 -0
  9. checkpoint-194000/rng_state_11.pth +3 -0
  10. checkpoint-194000/rng_state_12.pth +3 -0
  11. checkpoint-194000/rng_state_13.pth +3 -0
  12. checkpoint-194000/rng_state_14.pth +3 -0
  13. checkpoint-194000/rng_state_15.pth +3 -0
  14. checkpoint-194000/rng_state_2.pth +3 -0
  15. checkpoint-194000/rng_state_3.pth +3 -0
  16. checkpoint-194000/rng_state_4.pth +3 -0
  17. checkpoint-194000/rng_state_5.pth +3 -0
  18. checkpoint-194000/rng_state_6.pth +3 -0
  19. checkpoint-194000/rng_state_7.pth +3 -0
  20. checkpoint-194000/rng_state_8.pth +3 -0
  21. checkpoint-194000/rng_state_9.pth +3 -0
  22. checkpoint-194000/scheduler.pt +3 -0
  23. checkpoint-194000/trainer_state.json +3 -0
  24. checkpoint-194000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -75,3 +75,4 @@ checkpoint-186000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
75
  checkpoint-188000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-190000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
  checkpoint-192000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
75
  checkpoint-188000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-190000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
  checkpoint-192000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
+ checkpoint-194000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-194000/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-194000/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-194000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:124617666cee46918f6b1f25c2ef5cbdff126bd28af20c87afb56bb8604bfcaa
3
+ size 4539487822
checkpoint-194000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:597cc603f843d622b6dffdfb26075d2f7a6840e2bf8217c331c332c4646b5ce5
3
+ size 6632345482
checkpoint-194000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503a342257e35c7914f1721f01ebf64d88bcf913a51d73336ef52b1bf1c4dada
3
+ size 15984
checkpoint-194000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2d512040d5620cd810f06f7494a272f09a00ca4c7d15283a7f5f2b1f92563b6
3
+ size 15984
checkpoint-194000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:498172b51a908baec951c6482ec1856f1df4dbf0715728cf85810d95cc9c554e
3
+ size 15997
checkpoint-194000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34c95f4b314eeb798114d0f9c44920dabfc475faca4dbf8e741b77ff0b303a93
3
+ size 15997
checkpoint-194000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4de166eda0ead3a5c30e19488bc8b0ba551311e4a0f96939fc190553411195ac
3
+ size 15997
checkpoint-194000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21e2720fb06faee69c92b1c78064349fe373a932dc3fb2dc95b2f384e699b4b5
3
+ size 15997
checkpoint-194000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f505348a0077a50305ee9a58de41503bc8010ee022c017a4ad71b2fe737dc610
3
+ size 15997
checkpoint-194000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50ddbda597666e53c9255937715acd0eacb0e376b728f4831e363316c3b01da9
3
+ size 15997
checkpoint-194000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29af8dc5f8294f64f90a7a058e7050e3793502eaee55b19ac5c7ba823b574afe
3
+ size 15984
checkpoint-194000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a27095071f27903c5d8e9d3200d2c5828f647e752b172a7adf4b0ad5acbd781f
3
+ size 15984
checkpoint-194000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e857d3a03e92a4196253eb73d3850597a1c0de43fa5310f3e6f81c49ee8465b7
3
+ size 15984
checkpoint-194000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d30dd27db8a25dac77615bc8e7f909924f894134e38e12eb72217c45db21c0e5
3
+ size 15984
checkpoint-194000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33f4cfbd8999c12a9c7b3af496a8ba7199433d90f3caa8cc34702fc7f179fb31
3
+ size 15984
checkpoint-194000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c7bf6cf494b803a1864e881b64970d2b1248b441e34381e7937d13b446fe0ec
3
+ size 15984
checkpoint-194000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcd0866e7bef1817d19045eeb246bb58be0024a705906ee456e9f2a3e35eb800
3
+ size 15984
checkpoint-194000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dba572d5f74683cf2b63932c9eeb76b9f3578482f3bbf7c1b4e5bd3a613972d9
3
+ size 15984
checkpoint-194000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:517afd0d663592d538e5bd1d3a9acabff94a616c2a8c9867f6e7a70f0af30d8d
3
+ size 1064
checkpoint-194000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc3b1d6d97a63db0dfbd52f494a1f644f1594d05c3e27161dcfd7befd0930881
3
+ size 29996203
checkpoint-194000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ece67127b28907e8bbd26dc27ccc0ff7d44707e178dca4f8e7e2e4a2dc948c0d
3
+ size 6008