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
e36ff06
·
verified ·
1 Parent(s): 0b3a1ca

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-50000/config.json +79 -0
  2. checkpoint-50000/generation_config.json +6 -0
  3. checkpoint-50000/model.safetensors +3 -0
  4. checkpoint-50000/optimizer.pt +3 -0
  5. checkpoint-50000/rng_state_0.pth +3 -0
  6. checkpoint-50000/rng_state_1.pth +3 -0
  7. checkpoint-50000/rng_state_10.pth +3 -0
  8. checkpoint-50000/rng_state_11.pth +3 -0
  9. checkpoint-50000/rng_state_12.pth +3 -0
  10. checkpoint-50000/rng_state_13.pth +3 -0
  11. checkpoint-50000/rng_state_14.pth +3 -0
  12. checkpoint-50000/rng_state_15.pth +3 -0
  13. checkpoint-50000/rng_state_2.pth +3 -0
  14. checkpoint-50000/rng_state_3.pth +3 -0
  15. checkpoint-50000/rng_state_4.pth +3 -0
  16. checkpoint-50000/rng_state_5.pth +3 -0
  17. checkpoint-50000/rng_state_6.pth +3 -0
  18. checkpoint-50000/rng_state_7.pth +3 -0
  19. checkpoint-50000/rng_state_8.pth +3 -0
  20. checkpoint-50000/rng_state_9.pth +3 -0
  21. checkpoint-50000/scheduler.pt +3 -0
  22. checkpoint-50000/trainer_state.json +0 -0
  23. checkpoint-50000/training_args.bin +3 -0
checkpoint-50000/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-50000/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-50000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16cce835c8411863ed1a98cdf5da6b1012c9e061cb2a7fac739aef5cc37071ea
3
+ size 4539487822
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:929361e1f646225851972915125fadbee903affdc124eedc048f1949c2d02d5b
3
+ size 6632345482
checkpoint-50000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:852d0bf13991072bb2f4a9abfd3316bdb139e7703e2cf0623e4b40c9fa54f9ee
3
+ size 15984
checkpoint-50000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee069605483eb34cb89b9b3302146d2246614fdfce29eb551ab1dfbca0ff92fa
3
+ size 15984
checkpoint-50000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:297b302bc2821d24bd87b8c9cd3f739b1fec2f2fef6a57452bfaa6e665d6b9b2
3
+ size 15997
checkpoint-50000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c4cd3248fe60281d7ad308ac9bdfefa85d70e739d7d71e6cfd4b338b2f39594
3
+ size 15997
checkpoint-50000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec0624f7984b1ca3d3c452ec95177c28c894610e609d62c901176feeea93c6c0
3
+ size 15997
checkpoint-50000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a395d90533d740f3f3cb0662900bfb6449ea40a0baf06dc1e9e2bd9fe44bf41b
3
+ size 15997
checkpoint-50000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44a53da83b86f55a5a398d08bdbfb21925f4563204f3ebff97b747b907899de5
3
+ size 15997
checkpoint-50000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e510a1da1ee1430ff9bcb5d87d7e059cc9f326c5406fbb5b2489788df5dea67
3
+ size 15997
checkpoint-50000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1f6a8ad06e1e1059e002b30538843c3faa173cedf4bd39264db5739d8388ff0
3
+ size 15984
checkpoint-50000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d5b13f9bd4d94dfed37d7ad622b9a7ed749c3700abe3bc6fc288b1771c31336
3
+ size 15984
checkpoint-50000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:028d38d3ef3b9811a23db664896151b31dd873c5f9f606eb809a5648215c3295
3
+ size 15984
checkpoint-50000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5a05c1dbbd422b94352c5eda07c42ced40668495907c7329b8165d3b64dbde9
3
+ size 15984
checkpoint-50000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26396f2982a44f63c018b0945bf654b489c9e559b838b908bf67b7c9c793868b
3
+ size 15984
checkpoint-50000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c291b40efaa2dc81318f397c0d91f255f77d1f1115d402858f3d2ede3822d8a
3
+ size 15984
checkpoint-50000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5fb5f64b2b1844435aa4ef6084ea84f843454e65dd7afd129d09d20cf146824
3
+ size 15984
checkpoint-50000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1af6b157d6435c5946c6b538487274121b7ccbc794239dc29ab80688e769ab25
3
+ size 15984
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc44d1103e7d1616dd2d2f393a781b7690c295712b013d5889080915890b784f
3
+ size 1064
checkpoint-50000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce9fb82f83903b9d098b5f8d2a4ee340712cc3fa04639f01f1407c3beca3f19b
3
+ size 6008