https://github.com/alexjc/pytrain

🚃 Automated task/test framework for writing differentiable software.
https://github.com/alexjc/pytrain

deep-learning differentiable-programming python pytorch

Last synced: 10 months ago
JSON representation

🚃 Automated task/test framework for writing differentiable software.

• Host: GitHub
• URL: https://github.com/alexjc/pytrain
• Owner: alexjc
• License: bsd-3-clause
• Created: 2019-09-16T13:17:35.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2021-01-18T17:28:30.000Z (about 5 years ago)
• Last Synced: 2025-03-28T15:51:57.797Z (11 months ago)
• Topics: deep-learning, differentiable-programming, python, pytorch
• Language: Python
• Homepage:
• Size: 243 KB
• Stars: 7
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE

Awesome Lists containing this project

README

          
.. image:: docs/logo.png

.. image:: https://img.shields.io/badge/python-3.6%2B-blue

**What It Does.** Like ``pytest`` supports you in writing better Python code through automated testing, ``pytrain`` helps you build differentiable programs by making it easy to create tasks and use them to optimize your programs.

**How It Works.** You write differentiable components (e.g. using PyTorch) along with tasks they are expected to handle, and ``pytrain`` will optimize all the parameters involved and save them to disk as an automated process.

Installation

============

.. code:: bash

    # Create a base environment

    conda create -n myenv python=3.6

    conda install pytorch -c pytorch

    # Either install latest release 0.0.x, see GitHub for latest version number:

    pip install https://github.com/alexjc/pytrain/releases/download/v0.0.x/pytrain-0.0.x.tar.gz

    # Or clone the repository online:

    git clone https://github.com/alexjc/pytrain.git

Usage

=====

.. code:: bash

    # Either launch from installed script:

    pytrain -h

    pytrain --path examples/

    # Or run from current directory:

    python -m pytrain -h

    python -m pytrain --path examples/

Examples

========

See the ``#docs/`` folder or scripts in ``#examples/`` to get up and running.

NOTE: This version 0.0.x is an `early prototype `_ for the PyTorch Hackathon 2019.  Feedback and suggestions are the most welcome at this stage!

.. image:: docs/console.gif