Welcome to Jonas' CIFAR100 classifacation repo.
Features: Device agnostic training/testing platform for CIFAR100
Capable of saving models
Capable of loading models for training continuation or inference
Support for hyperparameter grid search
Exports data to Comet.ML for analysis
pip3 install comet_ml
install pytorch
From the command line: python ./train.py --help
Example: python ./train.py --model "linear" --save-path "./" --epochs 5
select from linear, conv, google models
python ./test.py --help python ./test.py --load-path "./path/to/saved/model"
shows test accuracy
Some experimental results shown in results subdirectory
All code contained in this project is my own or from the pytorch intro tutorial, unless specified otherwise.