Skip to content

A high level library on top of machine learning frameworks

License

Notifications You must be signed in to change notification settings

EKami/torchlite

Repository files navigation

Torchlite

PyPI version

Torchlite is a high level library on top of popular machine learning frameworks such as sklearn, Pytorch and Tensorflow. It gives a high layer abstraction of repetitive code used in machine learning for day-to-day data science tasks.

Installation

pip install torchlite

or if you want to run this lib directly to have access to the examples clone this repository and run:

pip install -r requirements.txt

to install the required dependencies. By default Pytorch 0.4.0+ and Tensorflow-GPU 1.8.0+ are installed along with this library but it's recommended to install them from source from here if you want to use the torchlite.torch package and/or head over to the Tensorflow install page if you want to use the torchlite.tf package.

Documentation

For now the library has no complete documentation but you can quickly get to know how it works by looking at the examples in the examples-* folders. This library is still in alpha and few APIs may change in the future. The only things which will evolve at the same pace as the library are the examples, they are meant to always be up to date with the library.

Few examples will generates folders/files such as saved models or tensorboard logs. To visualize the tensorboard logs download Tensorflow's tensorboard as well as Pytorch's tensorboard if you're interested by the torchlite.torch package. Then execute:

tensorboard --logdir=./tensorboard

Packaging the project for Pypi deploy

pip install twine
pip install wheel
python setup.py sdist
python setup.py bdist_wheel

Create a pypi account and create $HOME/.pypirc with:

[pypi]
username = <username>
password = <password>

Then upload the packages with:

twine upload dist/*

Or just:

pypi_deploy.sh

About

A high level library on top of machine learning frameworks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published