/fractionated-poolformer

Primary LanguagePythonMIT LicenseMIT

The idea

PoolFormer, but different channels have different pooling kernel sizes, and there's no downsampling.

Requirements

I develop inside of the January 2024 edition of the Nvidia PyTorch Docker image. docker run -it -d --gpus all -v /workspace:/workspace nvcr.io/nvidia/pytorch:24.01-py3

Repo structure

Implementations are in src, training script is in scripts along with a few sanity-checks. The training script expects CIFAR-10/100 to be in a folder called data.