ML Dependency Extras

In order to make it easy to make sure you are using dependencies which have been well tested with Beam ML, Beam provides a set of ML extras which can be installed alongside of Beam. For example, if you want to use a version of PyTorch which has been tested with Beam, you can install it with:

pip install beam[torch]

A full set of extras can be found in setup.py.

Note: You can also pin to dependencies outside of the extra range with a normal install - for example:

pip install beam==2.XX.0
pip install torch==<version released after Beam 2.XX.0>

this will usually work, but can break if the dependency releases a breaking change between the version Beam tests with and the version you pin to.