Hugging Face's logo

AiArtLab
/
sdxs3d

Diffusers
Safetensors
Model card Files
xet
Community
recoilme commited on
Commit
9cc90fc
·
1 Parent(s): b21699b

2409

Browse files
Files changed (10) hide show
  1. TRAIN.md → README.md +9 -8
  2. samples/unet_192x384_0.jpg +3 -0
  3. samples/unet_256x384_0.jpg +3 -0
  4. samples/unet_320x384_0.jpg +3 -0
  5. samples/unet_384x192_0.jpg +3 -0
  6. samples/unet_384x256_0.jpg +3 -0
  7. samples/unet_384x320_0.jpg +3 -0
  8. samples/unet_384x384_0.jpg +3 -0
  9. train.py +1 -1
  10. unet/diffusion_pytorch_model.safetensors +1 -1
TRAIN.md → README.md RENAMED
@@ -2,42 +2,43 @@
2
  license: apache-2.0
3
  ---
4
 
5
- Краткая инструкция по установке
6
- Обновите систему и установите git-lfs:
 
7
 
8
  ```
9
  apt update
10
  apt install git-lfs
11
  git config --global credential.helper store
12
  ```
13
- Обновите pip и установите требуемые пакеты:
14
 
15
  ```
16
  python -m pip install --upgrade pip
17
  pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu129 -U # cu129 - change
18
  pip install flash-attn --no-build-isolation # optional
19
  ```
20
- Клонируйте репозиторий:
21
 
22
  ```
23
  git clone https://huggingface.co/AiArtLab/sdxs3d
24
  cd sdxs3d/
25
  pip install -r requirements.txt -U
26
  ```
27
- Подготовьте датасет:
28
 
29
  ```
30
  mkdir datasets
31
  cd datasets
32
- huggingface-cli download AiArtLab/384 --local-dir 384 --repo-type dataset
33
  ```
34
- Выполните вход в сервисы:
35
 
36
  ```
37
  huggingface-cli login
38
  wandb login
39
  ```
40
- Запустите обучение!
41
 
42
  ```
43
  nohup accelerate launch train.py &
 
2
  license: apache-2.0
3
  ---
4
 
5
+ ## train
6
+
7
+ update:
8
 
9
  ```
10
  apt update
11
  apt install git-lfs
12
  git config --global credential.helper store
13
  ```
14
+ pip + torch:
15
 
16
  ```
17
  python -m pip install --upgrade pip
18
  pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu129 -U # cu129 - change
19
  pip install flash-attn --no-build-isolation # optional
20
  ```
21
+ repo:
22
 
23
  ```
24
  git clone https://huggingface.co/AiArtLab/sdxs3d
25
  cd sdxs3d/
26
  pip install -r requirements.txt -U
27
  ```
28
+ dataset:
29
 
30
  ```
31
  mkdir datasets
32
  cd datasets
33
+ huggingface-cli download AiArtLab/mjnj384 --local-dir mjnj --repo-type dataset
34
  ```
35
+ auth:
36
 
37
  ```
38
  huggingface-cli login
39
  wandb login
40
  ```
41
+ run it!
42
 
43
  ```
44
  nohup accelerate launch train.py &
samples/unet_192x384_0.jpg ADDED

Git LFS Details

  • SHA256: 8fa44929e0a212273a7740abd8ed37760709bc4eeb1cfdbadf5b703312880b6b
  • Pointer size: 130 Bytes
  • Size of remote file: 48.5 kB
samples/unet_256x384_0.jpg ADDED

Git LFS Details

  • SHA256: a98b63144b9008c744d824ba15031d80528cdd1a7009ef176d7f447281bf0a12
  • Pointer size: 130 Bytes
  • Size of remote file: 55.4 kB
samples/unet_320x384_0.jpg ADDED

Git LFS Details

  • SHA256: e6572072782819a52f85bb67f25d37f28131c22fa9d249b6d8d5e9bbb87da072
  • Pointer size: 130 Bytes
  • Size of remote file: 76.7 kB
samples/unet_384x192_0.jpg ADDED

Git LFS Details

  • SHA256: ba4d9ca34134987416b902d7c5058221b74676842ca425c46a177943a6140df3
  • Pointer size: 130 Bytes
  • Size of remote file: 38.4 kB
samples/unet_384x256_0.jpg ADDED

Git LFS Details

  • SHA256: 6597e59c9ababecd4d794bf0d22ced4d7999852dd6cf13e798208d91f27afbbe
  • Pointer size: 130 Bytes
  • Size of remote file: 54.8 kB
samples/unet_384x320_0.jpg ADDED

Git LFS Details

  • SHA256: 91f5733c86c545687de26950ac093ee63e5712da3ec50b0f051abf9e84bfd4bf
  • Pointer size: 130 Bytes
  • Size of remote file: 65.9 kB
samples/unet_384x384_0.jpg ADDED

Git LFS Details

  • SHA256: c6c9c166145efbbc1c318be2bc18d3bb1091e339f2549d7ba5409495bc89c70c
  • Pointer size: 130 Bytes
  • Size of remote file: 67.2 kB
train.py CHANGED
@@ -51,7 +51,7 @@ torch.backends.cuda.matmul.allow_tf32 = True
51
  torch.backends.cudnn.allow_tf32 = True
52
  torch.backends.cuda.enable_mem_efficient_sdp(False)
53
  dtype = torch.float32
54
- save_barrier = 1.001
55
  warmup_percent = 0.01
56
  percentile_clipping = 95 # 8bit optim
57
  betta2 = 0.97
 
51
  torch.backends.cudnn.allow_tf32 = True
52
  torch.backends.cuda.enable_mem_efficient_sdp(False)
53
  dtype = torch.float32
54
+ save_barrier = 1.05
55
  warmup_percent = 0.01
56
  percentile_clipping = 95 # 8bit optim
57
  betta2 = 0.97
unet/diffusion_pytorch_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:16fb28e82ad5c238b034ee615da6661457dd4151175085b46a3e48222c0d786a
3
  size 6184944280
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a1cab2d1280864a43c1fbba96bc6f29e344293ace5a004733a7bcd4a4e54c99
3
  size 6184944280