API Reference
=============

API reference for launching distributed GPU workloads and initializing Ray from Databricks notebooks.

distributed
-----------

.. currentmodule:: databricks.air

.. autofunction:: distributed

The decorator returns a :class:`~databricks.air.launcher.DistributedFunction` object.
Call ``.distributed()`` on it with the same arguments you would pass to your original
function to launch execution on GPU.

Accelerator Types
-----------------

.. currentmodule:: databricks.air.compute

.. note::

   ``accelerator_type`` is auto-detected from the notebook's attached accelerator. If you do specify it,
   you can pass the value as a string, such as ``"GPU_8xH100"``.

.. autoclass:: AcceleratorType
   :show-inheritance:
   :exclude-members: GPU_1xA10, GPU_1xH100, GPU_8xH100, GPU_8xB300, values, gpu_type, gpus_per_node, from_gpu_type

ray_init
--------

.. currentmodule:: databricks.air

.. autofunction:: ray_init