Implementation of diffusion in pure PyTorch.
-
Data - download from: https://www.kaggle.com/datasets/nunenuh/pytorch-challange-flower-dataset. Let
/path/to/data
be the path to the folder that contains thetrain
andtest
folders. -
Install:
conda create -n diffusion python=3.11 conda activate diffusion pip install -r requirements.txt
-
Training:
python main.py --conf conf.yml --command train --mnt-dir /path/to/data
See
conf.yml
for more options. -
Generate samples:
python main.py --conf conf.yml --command generate
See
conf.yml
for more options. -
Plot loss:
python main.py --conf conf.yml --command loss --show
-
Tests:
cd tests pytest