ray/doc/source/ray-air/config-scaling.rst

35 lines
1.1 KiB
ReStructuredText

.. _air-scaling-config:
Configuring Trainer Scaling
===========================
Ths guide describes how you can use the ``ScalingConfig`` object to configure resource utilization at the
per-run level when training models with Ray AIR.
Ray Train Usage
---------------
To use ``ScalingConfig`` when training a model, pass in the ``scaling_config`` parameter to your
``Trainer``:
.. literalinclude:: doc_code/config_scaling.py
:language: python
:start-after: __config_scaling_1__
:end-before: __config_scaling_1_end__
Ray Tune Usage
--------------
You can also treat some scaling config variables as hyperparameters and optimize them using Ray Tune.
Rather than passing in the ``scaling_config`` parameter to ``Trainer``, instead set the ``scaling_config``
key of the ``param_space`` dict that is passed to your ``Tuner`` initializer:
.. literalinclude:: doc_code/config_scaling.py
:language: python
:start-after: __config_scaling_2__
:end-before: __config_scaling_2_end__
For details on how Ray Tune resolves search spaces, see
:ref:`Ray Tune's search space tutorial <tune-search-space-tutorial>`.