Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
319eb68
·
verified ·
1 Parent(s): 97a9617

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-58000/config.json +79 -0
  2. checkpoint-58000/generation_config.json +6 -0
  3. checkpoint-58000/model.safetensors +3 -0
  4. checkpoint-58000/optimizer.pt +3 -0
  5. checkpoint-58000/rng_state_0.pth +3 -0
  6. checkpoint-58000/rng_state_1.pth +3 -0
  7. checkpoint-58000/rng_state_10.pth +3 -0
  8. checkpoint-58000/rng_state_11.pth +3 -0
  9. checkpoint-58000/rng_state_12.pth +3 -0
  10. checkpoint-58000/rng_state_13.pth +3 -0
  11. checkpoint-58000/rng_state_14.pth +3 -0
  12. checkpoint-58000/rng_state_15.pth +3 -0
  13. checkpoint-58000/rng_state_2.pth +3 -0
  14. checkpoint-58000/rng_state_3.pth +3 -0
  15. checkpoint-58000/rng_state_4.pth +3 -0
  16. checkpoint-58000/rng_state_5.pth +3 -0
  17. checkpoint-58000/rng_state_6.pth +3 -0
  18. checkpoint-58000/rng_state_7.pth +3 -0
  19. checkpoint-58000/rng_state_8.pth +3 -0
  20. checkpoint-58000/rng_state_9.pth +3 -0
  21. checkpoint-58000/scheduler.pt +3 -0
  22. checkpoint-58000/trainer_state.json +0 -0
  23. checkpoint-58000/training_args.bin +3 -0
checkpoint-58000/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": false,
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-58000/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-58000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10fbe7588cc6335164b38895df86e724c769df9171bac91d0795b36889445b1b
3
+ size 4538452374
checkpoint-58000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:784ed1ae1d91d981bbf5c5202ec80b3d88e06a534bc874cee69aa07759c65139
3
+ size 6630551906
checkpoint-58000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58e0295a0e995e2cddb25602506ce2f8509066b4addc372ab74b000462b02149
3
+ size 15984
checkpoint-58000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91f2ff2bcfc364e1382c42f2c1cf3bba139eeb1c06f7bf837c836a71b2d0bd30
3
+ size 15984
checkpoint-58000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e07626a898c6ad1d8182909fbd76cb9532e06a149e7840db095b255420101d60
3
+ size 15997
checkpoint-58000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d70ecf136da5456c540171062eb845c36bc6afb2f4f81fd04e7492b7468154b9
3
+ size 15997
checkpoint-58000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a97f144bd5ca2acb87e3c3b06cbc3845b30e9661024238dabb96d182c6a243c8
3
+ size 15997
checkpoint-58000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b60f15ee809aaef588d55935d52d214aa69be80f0826485b1566ad3392d24723
3
+ size 15997
checkpoint-58000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9d91fc17f519ab83fd782c4d200e90ad40b0a48877b9e0977e614eeed7af488
3
+ size 15997
checkpoint-58000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c77e8a7cca4176e29a36dd25a5f8bce426e4516ad08d7a85d3f379d45d3ae64
3
+ size 15997
checkpoint-58000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45e94f4bae01135f8d6aa8fc32c24232b09b1d2cb4aca77c5dcac6f5cedbb772
3
+ size 15984
checkpoint-58000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:426374c75b6a34104ffa9784914cd66b403900f400e6d650368f681b76eca83b
3
+ size 15984
checkpoint-58000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c184ec96ad26486618c229ecc2f7a0ad363fccf6af9c06e28301c9d720e2080
3
+ size 15984
checkpoint-58000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:094f59d5df1885bb9d02169fbc4fea53d59ecb7999f25ffafd444226aca04d1f
3
+ size 15984
checkpoint-58000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5446781e99a6cd7e260c5091c33dec4e81feb93932ee2b0d6970765fd2646885
3
+ size 15984
checkpoint-58000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07b425a239de3b1a28b0a505eeb4bc29e8667f25dcbbe204eeea32958bf0b7bc
3
+ size 15984
checkpoint-58000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ee30b846da3e446f9ec177874be163fd7f20a93a537c1dca1c4d267875cb261
3
+ size 15984
checkpoint-58000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75745ba7f53ce2b5f7d5363c9fb891d6b902a6d664dbea1c83a335528800f17d
3
+ size 15984
checkpoint-58000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:874dfffb78647a5518ae4417a9ac95605ec9fb67f59539b5a10b609d82e13ded
3
+ size 1064
checkpoint-58000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-58000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3775d3fa592932a55430eaf3bead369ffe2032e782d1e335a7c29ce2e11cbe3e
3
+ size 5944