Installation
============

Requirements
------------

PyTorch Lightning requires:

* Python 3.9 or later
* PyTorch 2.0 or later
* CUDA 11.8+ (optional, for GPU acceleration)

Install from PyPI
-----------------

The recommended way to install PyTorch Lightning:

.. code-block:: bash

   pip install lightning

This installs the unified ``lightning`` package which includes ``lightning.pytorch``.
To install the standalone PyTorch Lightning package:

.. code-block:: bash

   pip install pytorch-lightning

Install with extras
~~~~~~~~~~~~~~~~~~~

.. code-block:: bash

   # Weights & Biases integration
   pip install "lightning[wandb]"

   # MLflow integration
   pip install "lightning[mlflow]"

   # All optional integrations
   pip install "lightning[extra]"

Install from source
-------------------

.. code-block:: bash

   git clone https://github.com/Lightning-AI/pytorch-lightning.git
   cd pytorch-lightning
   pip install -e ".[dev]"

Verifying installation
----------------------

.. code-block:: python

   import lightning as L
   print(L.__version__)  # 2.4.0
   import lightning.pytorch as pl
   print(pl.__version__)  # 2.4.0

Docker
------

An official Docker image is available:

.. code-block:: bash

   docker pull pytorchlightning/pytorch_lightning:base-cuda-py3.11-torch2.4
   docker run --gpus all -it pytorchlightning/pytorch_lightning:base-cuda-py3.11-torch2.4

.. note::

   When using Docker with GPU support, ensure the NVIDIA Container Toolkit is installed
   on the host machine.