Pytorch Lighning Classification

Description

• Train any torchvision model on any torchvision dataset.
• Use Pytorch Lighning to reduce code size.
• Create Callbacks to split every part of Deep Learning models and methods.

How to use

Prepare environment

Install python >= 3.9:

conda create -n plcls python=3.9 -y
conda deactivate && conda activate plcls

Start training

Yaml config is all you need:

python main.py configs/config.yaml

Get results

See lightning_logs/version_0/ folder for training results.

See lightning_logs/version_0/tb/ for tensorboard plots:

tensorboard --logdir lightning_logs/

Roadmap

• Training code
• Evaluation
• Custom datasets
• Custom models
• SAM, more augs, etc. (efficient training)
• Tests, linters, pipelines