2022-06-15 08:51:37 -06:00
|
|
|
## Training examples
|
|
|
|
|
|
|
|
|
|
### Flowers DDPM
|
|
|
|
|
|
|
|
|
|
The command to train a DDPM UNet model on the Oxford Flowers dataset:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
python -m torch.distributed.launch \
|
|
|
|
|
--nproc_per_node 4 \
|
|
|
|
|
train_ddpm.py \
|
|
|
|
|
--dataset="huggan/flowers-102-categories" \
|
|
|
|
|
--resolution=64 \
|
|
|
|
|
--output_path="flowers-ddpm" \
|
|
|
|
|
--batch_size=16 \
|
|
|
|
|
--num_epochs=100 \
|
|
|
|
|
--gradient_accumulation_steps=1 \
|
|
|
|
|
--lr=1e-4 \
|
|
|
|
|
--warmup_steps=500 \
|
|
|
|
|
--mixed_precision=no
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
A full ltraining run takes 2 hours on 4xV100 GPUs.
|
|
|
|
|
|
|
|
|
|
<img src="https://user-images.githubusercontent.com/26864830/173855866-5628989f-856b-4725-a944-d6c09490b2df.png" width="500" />
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
### Pokemon DDPM
|
|
|
|
|
|
|
|
|
|
The command to train a DDPM UNet model on the Pokemon dataset:
|
|
|
|
|
|
|
|
|
|
```bash
|
|
|
|
|
python -m torch.distributed.launch \
|
|
|
|
|
--nproc_per_node 4 \
|
|
|
|
|
train_ddpm.py \
|
|
|
|
|
--dataset="huggan/pokemon" \
|
|
|
|
|
--resolution=64 \
|
2022-06-15 13:45:49 -06:00
|
|
|
--output_path="pokemon-ddpm" \
|
2022-06-15 08:51:37 -06:00
|
|
|
--batch_size=16 \
|
|
|
|
|
--num_epochs=100 \
|
|
|
|
|
--gradient_accumulation_steps=1 \
|
|
|
|
|
--lr=1e-4 \
|
|
|
|
|
--warmup_steps=500 \
|
|
|
|
|
--mixed_precision=no
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
A full ltraining run takes 2 hours on 4xV100 GPUs.
|
|
|
|
|
|
|
|
|
|
<img src="https://user-images.githubusercontent.com/26864830/173856733-4f117f8c-97bd-4f51-8002-56b488c96df9.png" width="500" />
|
