A Cookiecutter template for PyTorch projects.

• Python >= 3.8
• PyTorch >= 1.1
• Tensorboard >= 1.4

• Clear folder structure which is suitable for many deep learning projects.
• Runs are configured via .yml files allowing for easy experimentation.
• Checkpoint saving and resuming.
• Tensorboard logging

$ pip install cookiecutter
$ cookiecutter https://github.com/khornlund/cookiecutter-pytorch
$ cd path/to/repo

A template project has now been created! You can run the MNIST example using:

$ conda env create --file environment.yml
$ conda activate <your-project-name>
$ <your-project-name> train

Here are some projects which use this template:

1. Severstal Steel Defect Detection (Kaggle)
2. Aptos Blindness Detection (Kaggle)
3. Understanding Cloud Organization (Kaggle)

If you fork this repo, you can modify cookiecutter.json to provide personalised defaults eg. name, email, username, etc.

This template was based on PyTorch Template.