audio-diffusion / README.md
teticio's picture
update for sagemaker
65fa65c
|
raw
history blame
474 Bytes
# audio-diffusion
```bash
accelerate config
```
```bash
python src/audio_to_images.py \
--resolution 256 \
--input_dir path-to-audio-files \
--output_dir data-256
```
```bash
accelerate launch src/train_unconditional.py \
--dataset_name data-256 \
--resolution 256 \
--output_dir ddpm-ema-audio-256 \
--train_batch_size 16 \
--num_epochs 100 \
--gradient_accumulation_steps 1 \
--learning_rate 1e-4 \
--lr_warmup_steps 500 \
--mixed_precision no
```