https://github.com/oscarknagg/olympic-pytorch
Python package implementing a Keras-like API for Pytorch
https://github.com/oscarknagg/olympic-pytorch
Last synced: over 1 year ago
JSON representation
Python package implementing a Keras-like API for Pytorch
- Host: GitHub
- URL: https://github.com/oscarknagg/olympic-pytorch
- Owner: oscarknagg
- License: mit
- Created: 2018-11-11T16:51:26.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2022-12-08T01:27:26.000Z (over 3 years ago)
- Last Synced: 2025-03-21T15:12:26.082Z (over 1 year ago)
- Language: Python
- Size: 62.5 KB
- Stars: 5
- Watchers: 2
- Forks: 0
- Open Issues: 13
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Olympic PyTorch
Olympic implements a Keras-like API for PyTorch.
The goal of Olympic is to combine the joy of Pytorch's dynamic graph execution with the joy of Keras's high level abstractions
for training. Concretely, Olympic contains:
1. The ``olympic.fit()`` function. This implements a very similar API
to Keras's ``model.fit`` and ``model.fit_generator`` methods in a
more functional and less object-oriented fashion and spares you
the effort of "hand-rolling" your own training loop.
2. ``Callback`` objects that perform functionality common to most deep
learning training pipelines such as learning rate scheduling, model
checkpointing and csv logging. These integrate into ``olympic.fit()``
and spare you the effort of writing boilerplate code.
3. Some helpful utility functions such as common metrics and some
convenience layers from Keras that are missing in PyTorch.
Documentation is available [here](https://olympic-pytorch.readthedocs.io/en/latest/)
This package is available on PyPi.
```
pip install olympic
```