Small private project to get used to the Pytorch framework. It is meant to have somewhat of a modular structure to support testing of different models depending on input args, predefined models and varying datasets.

• /src source code with all .py files. Check the information inside this folder for a closer look.

• /models .pth files of the trained models. Stores the final weights of model. Use torch.load() to load it back in and make predictions on custom images (see src/apply.py)

• /documentation some images of the dataset and wrong guesses plus ground truth

• /data placeholder folder for downloaded datasets. Additionally stores custom drawn numbers to test real life examples.

python3 -m venv env
source env/bin/activate
pip install -r requirements.txt
python3 src/main.py --model 3 --epochs 5 --optimizer 3 --seed 42 --filename myModel