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
7d91fee
·
verified ·
1 Parent(s): 755ba88

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-84000/config.json +79 -0
  3. checkpoint-84000/generation_config.json +6 -0
  4. checkpoint-84000/model.safetensors +3 -0
  5. checkpoint-84000/optimizer.pt +3 -0
  6. checkpoint-84000/rng_state_0.pth +3 -0
  7. checkpoint-84000/rng_state_1.pth +3 -0
  8. checkpoint-84000/rng_state_10.pth +3 -0
  9. checkpoint-84000/rng_state_11.pth +3 -0
  10. checkpoint-84000/rng_state_12.pth +3 -0
  11. checkpoint-84000/rng_state_13.pth +3 -0
  12. checkpoint-84000/rng_state_14.pth +3 -0
  13. checkpoint-84000/rng_state_15.pth +3 -0
  14. checkpoint-84000/rng_state_2.pth +3 -0
  15. checkpoint-84000/rng_state_3.pth +3 -0
  16. checkpoint-84000/rng_state_4.pth +3 -0
  17. checkpoint-84000/rng_state_5.pth +3 -0
  18. checkpoint-84000/rng_state_6.pth +3 -0
  19. checkpoint-84000/rng_state_7.pth +3 -0
  20. checkpoint-84000/rng_state_8.pth +3 -0
  21. checkpoint-84000/rng_state_9.pth +3 -0
  22. checkpoint-84000/scheduler.pt +3 -0
  23. checkpoint-84000/trainer_state.json +3 -0
  24. checkpoint-84000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -38,3 +38,4 @@ checkpoint-76000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
  checkpoint-78000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
40
  checkpoint-82000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
38
  checkpoint-78000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
40
  checkpoint-82000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
+ checkpoint-84000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-84000/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-84000/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-84000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12b71ccf545240f6aacc4af8a92ad527d573e25b48c0efd0cd7a2d50873cfd86
3
+ size 4539487822
checkpoint-84000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33b97083c7edbf565ea5bcf6963191c711c066f73750c8997f8c12b67e81afdd
3
+ size 6632345482
checkpoint-84000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8aea7eb4a817c8efb35a34bf159da5bf07e302e89b19ba6b0791f94d6cb9a784
3
+ size 15984
checkpoint-84000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0be7bde091d52f139aa9d885dccb64dec7b86feddeaed344ebec0214eb8801a5
3
+ size 15984
checkpoint-84000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66584c0396300654ed6ea21769b9415ccb790652fb4bb4ee22f53fa3b1df6e90
3
+ size 15997
checkpoint-84000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04d3845c83c389b16fd352b9bc5f6da3a73f3480381357d483cd1297053e24a3
3
+ size 15997
checkpoint-84000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cf81544a21e46b213cae02707552a54a3ead87f182c35ed76bb05e100942a4d
3
+ size 15997
checkpoint-84000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a430d0b12b1fd4be151513afdc2a41247cee32e2b223bf5444c617e64774d33
3
+ size 15997
checkpoint-84000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50a5c99cd005d997a41ff70791628029b1311bc04086fc279b39cc1aeeb9ac2e
3
+ size 15997
checkpoint-84000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3076269ed5e4efa9a2c696a28a9db74bd3edf29ff9379731b9f9ec69bb48350c
3
+ size 15997
checkpoint-84000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed66a29e083adae97a215f27713ee1581844864bc574c88986979f0589b5a65c
3
+ size 15984
checkpoint-84000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58b3759a6a0c392469cf8cc84d535e520b05ed60d67b1a0bd2d57a74aedf73b9
3
+ size 15984
checkpoint-84000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b448f2322a45797120c7d3732958d69d978184ee9a7c1ea73ebc1aa5197fd1bb
3
+ size 15984
checkpoint-84000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdaf29e678a2efb52b41a876311084eb29905b290325f4f3d0ad1a4c92a6fe62
3
+ size 15984
checkpoint-84000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00d316dac63180fefa5fedd34f05dc12ca627a3da90f37c52200f34c33b75b24
3
+ size 15984
checkpoint-84000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:177d772f0735ea06fd6970db4dbf3a8cd34737d0122e43f9b5552e51fdce0957
3
+ size 15984
checkpoint-84000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fac18d9c7392e790ecc07af6d4f4bd3c0b92da4cc0fac743d67b60e66eb7896
3
+ size 15984
checkpoint-84000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:161ca537261e2b2ff48588fe112bb1fc834417096b77ad6db8077d82c1fb6593
3
+ size 15984
checkpoint-84000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00f3e54d4d00df9d70be0ad429999f002970ffa47dc120caa74fba0cec12cf46
3
+ size 1064
checkpoint-84000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f4114105dd2faadd3e51679027a76e2da1282ee967028196a497ec975164313
3
+ size 12954989
checkpoint-84000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d790f7d8e5ad5da9de24ef5c64cadbf92142701e9edd51bd954eb1f6922ad6b
3
+ size 6008