audio-diffusion / README.md
teticio's picture
update for sagemaker
65fa65c
|
raw
history blame
474 Bytes

audio-diffusion

accelerate config
python src/audio_to_images.py \
  --resolution 256 \
  --input_dir path-to-audio-files \
  --output_dir data-256
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