This repo contains simple reimplementation of Flow Matching paper: Lipman, Yaron, et al
To train our model we used MNIST dataset.
pip install -r requirements.txt
python src/train.py
There is a couple of settings you may want to specify:
--batch_size- set depending on your gpu memory available--num_epochs- num epoch to train the model--lr- learning rate--device- which device to use--output_path- path to save training artefacts
python src/inference.py
There is a couple of settings you may want to specify:
--checkpoint_filepath- path to pretrained model--num_samples- how many samples to generate--device- which device to use--output_path- filepath to save result image
Generated samples