Fast Autoaugment: Search image augmentation policies

Understand the project

Practical overview

Use it when you want to explore learned image augmentation within a PyTorch training workflow and can configure the documented search infrastructure.

Computer-vision researchers and developers using PyTorch image-classification pipelines.
Practitioners comparing learned augmentation policies with baseline or Cutout training.

Search for image augmentation policies with a configured Ray cluster.
Train models using searched policies on documented datasets such as CIFAR-10, CIFAR-100 and ImageNet.
Compare direct or transferred policy results with the README's reported baselines.

Before you choose

The public GitHub repository identifies the project as MIT-licensed.

Policy search depends on a properly configured Ray cluster and a Redis address.
The README's documented experiment environment is Python 3.6.9, PyTorch 1.2.0, torchvision 0.4.0 and CUDA 10.
The repository provides separate search and training stages; training examples use particular policy names, datasets and model configurations.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Launch a documented CIFAR-10 or CIFAR-100 training command with the `fa_reduced_cifar10` policy.

Add the repository working directory to PYTHONPATH with export PYTHONPATH=$PYTHONPATH:$PWD.
Choose a documented configuration, such as confs/wresnet40x2_cifar10_b512.yaml.

Run the command below, substituting cifar100 for the CIFAR-100 dataset.

Documented command

python FastAutoAugment/train.py -c confs/wresnet40x2_cifar10_b512.yaml --aug fa_reduced_cifar10 --dataset cifar10

The training script is invoked with the selected configuration, policy and dataset.

The repository identifies this policy as based on reduced CIFAR-10 (4,000 images) and WResNet-40x2; the examples do not establish results for other configurations.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 16 October 2026
Original website title: kakaobrain/fast-autoaugment · Official homepage title · retained in its source language