Even using frameworks such as Keras or Pytorch Lightning leaves you with making many choices on how to organize your machine learning project. The main ambition of this repository is to make easy following the state-of-the-art good practices for a machine learning project, for the most popular packages/frameworks.
Maintained and distributed by GMUM (https://gmum.net/).
The following templates are provided:
-
pytorch_lightning_project_template - template for a ML project based on PyTorch and Pytorch Lightning. Advocates using modular code, gin configuration, and neptune. Last updated 10.2020.
-
tf2_project_template - template for a ML project based on vanilla TF2. Last updated 6.2020. Includes generic improvements compared to the PyTorch template such as automatic syncing of files, using watchman or integration with Neptune, or standalone training loop.
-
pytorch_project_template - template for a ML project based on vanilla pytorch. Last updated 12.2019.
-
keras_project_template - template for a ML project based on keras. Last updated 04.2018.