pytorch-lightning Documentation

Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.

31,283 stars Python Python 1 file · ~7,204 tokens #ai#artificial-intelligence#data-science#deep-learning#machine-learning#python#pytorch
RAW Doc

File: README.md

 

Why PyTorch Lightning?

Training models in plain PyTorch requires writing and maintaining a lot of repetitive engineering code. Handling backpropagation, mixed precision, multi-GPU, and distributed training is error-prone and often reimplemented for every project. PyTorch Lightning organizes PyTorch code to automate this infrastructure while keeping full control over your model logic. You write the science. Lightning handles the engineering, and scales from CPU to multi-node GPUs without changing your core code. PyTorch experts can still opt into expert-level control.

Fun analogy: If PyTorch is Javascript, PyTorch Lightning is ReactJS or NextJS.

Looking for GPUs?

Lightning Cloud is the easiest way to run PyTorch Lightning without managing infrastructure. Start training with one command and get GPUs, autoscaling, monitoring, and a free tier. No cloud setup required.

You can also run PyTorch Lightning on your own hardware or cloud.

Lightning has 2 core packages

PyTorch Lightning: Train and deploy PyTorch at scale.

Lightning Fabric: Expert control.

Lightning gives you granular control over how much abstraction you want to add over PyTorch.

 

Quick start

Install Lightning:

bash
pip install lightning

Advanced install options

Install with optional dependencies

bash
pip install lightning['extra']

Conda

bash
conda install lightning -c conda-forge

Install stable version

Install future release from the source

bash
pip install https://github.com/Lightning-AI/lightning/archive/refs/heads/release/stable.zip -U

Install bleeding-edge

Install nightly from the source (no guarantees)

bash
pip install https://github.com/Lightning-AI/lightning/archive/refs/heads/master.zip -U

or from testing PyPI

bash
pip install -iU https://test.pypi.org/simple/ pytorch-lightning

PyTorch Lightning example

Define the training workflow. Here's a toy example (explore real examples):

python
# main.py
# ! pip install torchvision
import torch, torch.nn as nn, torch.utils.data as data, torchvision as tv, torch.nn.functional as F
import lightning as L

# --------------------------------
# Step 1: Define a LightningModule
# --------------------------------
# A LightningModule (nn.Module subclass) defines a full *system*
# (ie: an LLM, diffusion model, autoencoder, or simple image classifier).


class LitAutoEncoder(L.LightningModule):
    def __init__(self):
        super().__init__()
        self.encoder = nn.Sequential(nn.Linear(28 * 28, 128), nn.ReLU(), nn.Linear(128, 3))
        self.decoder = nn.Sequential(nn.Linear(3, 128), nn.ReLU(), nn.Linear(128, 28 * 28))

    def forward(self, x):
        # in lightning, forward defines the prediction/inference actions
        embedding = self.encoder(x)
        return embedding

    def training_step(self, batch, batch_idx):
        # training_step defines the train loop. It is independent of forward
        x, _ = batch
        x = x.view(x.size(0), -1)
        z = self.encoder(x)
        x_hat = self.decoder(z)
        loss = F.mse_loss(x_hat, x)
        self.log("train_loss", loss)
        return loss

    def configure_optimizers(self):
        optimizer = torch.optim.Adam(self.parameters(), lr=1e-3)
        return optimizer


# -------------------
# Step 2: Define data
# -------------------
dataset = tv.datasets.MNIST(".", download=True, transform=tv.transforms.ToTensor())
train, val = data.random_split(dataset, [55000, 5000])

# -------------------
# Step 3: Train
# -------------------
autoencoder = LitAutoEncoder()
trainer = L.Trainer()
trainer.fit(autoencoder, data.DataLoader(train), data.DataLoader(val))

Run the model on your terminal

bash
pip install torchvision
python main.py

 

Convert from PyTorch to PyTorch Lightning

PyTorch Lightning is just organized PyTorch - Lightning disentangles PyTorch code to decouple the science from the engineering.

 

Examples

Explore various types of training possible with PyTorch Lightning. Pretrain and finetune ANY kind of model to perform ANY task like classification, segmentation, summarization and more:

Task Description Run
Hello world Pretrain - Hello world example
Image classification Finetune - ResNet-34 model to classify images of cars
Image segmentation Finetune - ResNet-50 model to segment images
Object detection Finetune - Faster R-CNN model to detect objects
Text classification Finetune - text classifier (BERT model)
Text summarization Finetune - text summarization (Hugging Face transformer model)
Audio generation Finetune - audio generator (transformer model)
LLM finetuning Finetune - LLM (Meta Llama 3.1 8B)
Image generation Pretrain - Image generator (diffusion model)
Recommendation system Train - recommendation system (factorization and embedding)
Time-series forecasting Train - Time-series forecasting with LSTM

Advanced features

Lightning has over 40+ advanced features
designed for professional AI research at scale.

Here are some examples:

Train on 1000s of GPUs without code changes

python
# 8 GPUs
# no code changes needed
trainer = Trainer(accelerator="gpu", devices=8)

# 256 GPUs
trainer = Trainer(accelerator="gpu", devices=8, num_nodes=32)

Train on other accelerators like TPUs without code changes

python
# no code changes needed
trainer = Trainer(accelerator="tpu", devices=8)

16-bit precision

python
# no code changes needed
trainer = Trainer(precision=16)

Experiment managers

python
from lightning import loggers

# litlogger
trainer = Trainer(logger=LitLogger())

# tensorboard
trainer = Trainer(logger=TensorBoardLogger("logs/"))

# weights and biases
trainer = Trainer(logger=loggers.WandbLogger())

# comet
trainer = Trainer(logger=loggers.CometLogger())

# mlflow
trainer = Trainer(logger=loggers.MLFlowLogger())

# ... and dozens more

Early Stopping

python
es = EarlyStopping(monitor="val_loss")
trainer = Trainer(callbacks=[es])

Checkpointing

python
checkpointing = ModelCheckpoint(monitor="val_loss")
trainer = Trainer(callbacks=[checkpointing])

Export to torchscript (JIT) (production use)

python
# torchscript
autoencoder = LitAutoEncoder()
torch.jit.save(autoencoder.to_torchscript(), "model.pt")

Export to ONNX (production use)

python
# onnx
with tempfile.NamedTemporaryFile(suffix=".onnx", delete=False) as tmpfile:
    autoencoder = LitAutoEncoder()
    input_sample = torch.randn((1, 64))
    autoencoder.to_onnx(tmpfile.name, input_sample, export_params=True)
    os.path.isfile(tmpfile.name)

Advantages over unstructured PyTorch

  • Models become hardware agnostic
  • Code is clear to read because engineering code is abstracted away
  • Easier to reproduce
  • Make fewer mistakes because lightning handles the tricky engineering
  • Keeps all the flexibility (LightningModules are still PyTorch modules), but removes a ton of boilerplate
  • Lightning has dozens of integrations with popular machine learning tools.
  • Tested rigorously with every new PR. We test every combination of PyTorch and Python supported versions, every OS, multi GPUs and even TPUs.
  • Minimal running speed overhead (about 300 ms per epoch compared with pure PyTorch).

 
 

Lightning Fabric: Expert control

Run on any device at any scale with expert-level control over PyTorch training loop and scaling strategy. You can even write your own Trainer.

Fabric is designed for the most complex models like foundation model scaling, LLMs, diffusion, transformers, reinforcement learning, active learning. Of any size.

What to change Resulting Fabric Code (copy me!)
```diff
  • import lightning as L
    import torch; import torchvision as tv

dataset = tv.datasets.CIFAR10("data", download=True,
train=True,
transform=tv.transforms.ToTensor())

  • fabric = L.Fabric()

  • fabric.launch()

    model = tv.models.resnet18()
    optimizer = torch.optim.SGD(model.parameters(), lr=0.001)

  • device = "cuda" if torch.cuda.is_available() else "cpu"
  • model.to(device)

  • model, optimizer = fabric.setup(model, optimizer)

    dataloader = torch.utils.data.DataLoader(dataset, batch_size=8)

  • dataloader = fabric.setup_dataloaders(dataloader)

    model.train()
    num_epochs = 10
    for epoch in range(num_epochs):
    for batch in dataloader:
    inputs, labels = batch

  • text
    inputs, labels = inputs.to(device), labels.to(device)
    optimizer.zero_grad()
    outputs = model(inputs)
    loss = torch.nn.functional.cross_entropy(outputs, labels)
  • text
    loss.backward()
  • text
    fabric.backward(loss)
    optimizer.step()
    print(loss.data)
text
```Python
import lightning as L
import torch; import torchvision as tv

dataset = tv.datasets.CIFAR10("data", download=True,
                              train=True,
                              transform=tv.transforms.ToTensor())

fabric = L.Fabric()
fabric.launch()

model = tv.models.resnet18()
optimizer = torch.optim.SGD(model.parameters(), lr=0.001)
model, optimizer = fabric.setup(model, optimizer)

dataloader = torch.utils.data.DataLoader(dataset, batch_size=8)
dataloader = fabric.setup_dataloaders(dataloader)

model.train()
num_epochs = 10
for epoch in range(num_epochs):
    for batch in dataloader:
        inputs, labels = batch
        optimizer.zero_grad()
        outputs = model(inputs)
        loss = torch.nn.functional.cross_entropy(outputs, labels)
        fabric.backward(loss)
        optimizer.step()
        print(loss.data)
``` |  |



## Key features


  Easily switch from running on CPU to GPU (Apple Silicon, CUDA, …), TPU, multi-GPU or even multi-node training

```python
# Use your available hardware
# no code changes needed
fabric = Fabric()

# Run on GPUs (CUDA or MPS)
fabric = Fabric(accelerator="gpu")

# 8 GPUs
fabric = Fabric(accelerator="gpu", devices=8)

# 256 GPUs, multi-node
fabric = Fabric(accelerator="gpu", devices=8, num_nodes=32)

# Run on TPUs
fabric = Fabric(accelerator="tpu")

Use state-of-the-art distributed training strategies (DDP, FSDP, DeepSpeed) and mixed precision out of the box

python
# Use state-of-the-art distributed training techniques
fabric = Fabric(strategy="ddp")
fabric = Fabric(strategy="deepspeed")
fabric = Fabric(strategy="fsdp")

# Switch the precision
fabric = Fabric(precision="16-mixed")
fabric = Fabric(precision="64")

All the device logic boilerplate is handled for you

diff
# no more of this!
- model.to(device)
- batch.to(device)

Build your own custom Trainer using Fabric primitives for training checkpointing, logging, and more

python
import lightning as L


class MyCustomTrainer:
    def __init__(self, accelerator="auto", strategy="auto", devices="auto", precision="32-true"):
        self.fabric = L.Fabric(accelerator=accelerator, strategy=strategy, devices=devices, precision=precision)

    def fit(self, model, optimizer, dataloader, max_epochs):
        self.fabric.launch()

        model, optimizer = self.fabric.setup(model, optimizer)
        dataloader = self.fabric.setup_dataloaders(dataloader)
        model.train()

        for epoch in range(max_epochs):
            for batch in dataloader:
                input, target = batch
                optimizer.zero_grad()
                output = model(input)
                loss = loss_fn(output, target)
                self.fabric.backward(loss)
                optimizer.step()

You can find a more extensive example in our examples

 
 

Examples

Self-supervised Learning

Convolutional Architectures

Reinforcement Learning

GANs

Classic ML

 
 

Continuous Integration

Lightning is rigorously tested across multiple CPUs, GPUs and TPUs and against major Python and PyTorch versions.

\*Codecov is > 90%+ but build delays may show less

Current build statuses

System / PyTorch ver. 1.13 2.0 2.1
Linux py3.9 [GPUs]
Linux (multiple Python versions)
OSX (multiple Python versions)
Windows (multiple Python versions)

 
 

Community

The lightning community is maintained by

  • 10+ core contributors who are all a mix of professional engineers, Research Scientists, and Ph.D. students from top AI labs.
  • 800+ community contributors.

Want to help us build Lightning and reduce boilerplate for thousands of researchers? Learn how to make your first contribution here

Lightning is also part of the PyTorch ecosystem which requires projects to have solid testing, documentation and support.

Asking for help

If you have any questions please:

  1. Read the docs.
  2. Search through existing Discussions, or add a new question
  3. Join our discord.

File: dockers/README.md

Docker images

Build images from Dockerfiles

You can build it on your own, note it takes lots of time, be prepared.

bash
git clone https://github.com/Lightning-AI/lightning.git

# build with the default arguments
docker image build -t pytorch-lightning:latest -f dockers/base-cuda/Dockerfile .

# build with specific arguments
docker image build -t pytorch-lightning:base-cuda-py3.9-torch1.13-cuda11.7.1 -f dockers/base-cuda/Dockerfile --build-arg PYTHON_VERSION=3.9 --build-arg PYTORCH_VERSION=1.13 --build-arg CUDA_VERSION=11.7.1 .

To run your docker use

bash
docker image list
docker run --rm -it pytorch-lightning:latest bash

and if you do not need it anymore, just clean it:

bash
docker image list
docker image rm pytorch-lightning:latest

Run docker image with GPUs

To run docker image with access to your GPUs, you need to install

bash
# Add the package repositories
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

and later run the docker image with --gpus all. For example,

text
docker run --rm -it --gpus all pytorchlightning/pytorch_lightning:base-cuda-py3.9-torch1.13-cuda11.7.1

Run Jupyter server

  1. Build the docker image:
    bash
    docker image build -t pytorch-lightning:v1.6.5 -f dockers/nvidia/Dockerfile --build-arg LIGHTNING_VERSION=1.6.5 .
  2. start the server and map ports:
    bash
    docker run --rm -it --gpus=all -p 8888:8888 pytorch-lightning:v1.6.5
  3. Connect in local browser:
    • copy the generated path e.g. http://hostname:8888/?token=0719fa7e1729778b0cec363541a608d5003e26d4910983c6
    • replace the hostname by localhost

File: docs/source-app/api_reference/components.rst

########################
lightning.app.components
########################

.. contents::
:depth: 1
:local:
:backlinks: top

.. currentmodule:: lightning.app.components

Built-in Components

.. autosummary::
:toctree: generated/
:nosignatures:
:template: classtemplate.rst

text
~database.client.DatabaseClient
~database.server.Database
~python.popen.PopenPythonScript
~python.tracer.TracerPythonScript
~training.LightningTrainerScript
~serve.gradio_server.ServeGradio
~serve.serve.ModelInferenceAPI
~serve.python_server.PythonServer
~serve.streamlit.ServeStreamlit
~multi_node.base.MultiNode
~multi_node.fabric.FabricMultiNode
~multi_node.pytorch_spawn.PyTorchSpawnMultiNode
~multi_node.trainer.LightningTrainerMultiNode
~serve.auto_scaler.AutoScaler
~serve.auto_scaler.ColdStartProxy

File: docs/source-app/api_reference/core.rst

:orphan:

##################
lightning.app.core
##################

.. contents::
:depth: 1
:local:
:backlinks: top

.. currentmodule:: lightning.app.core

Core APIs

.. autosummary::
:toctree: api/
:nosignatures:
:template: classtemplate.rst

text
LightningApp
LightningFlow
LightningWork

Learn more about :ref:Lightning Core <core_api>.

File: docs/source-app/api_reference/frontend.rst

######################
lightning.app.frontend
######################

.. contents::
:depth: 1
:local:
:backlinks: top

.. currentmodule:: lightning.app.frontend

Lightning FrontEnds

.. autosummary::
:toctree: generated/
:nosignatures:
:template: classtemplate.rst

text
~frontend.Frontend
~web.StaticWebFrontend
~stream_lit.StreamlitFrontend
~panel.PanelFrontend

Learn more about :ref:Frontend's .

File: docs/source-app/api_reference/runners.rst

#####################
lightning.app.runners
#####################

.. contents::
:depth: 1
:local:
:backlinks: top

.. currentmodule:: lightning.app.runners

Lightning Core

.. autosummary::
:toctree: generated/
:nosignatures:
:template: classtemplate.rst

text
~cloud.CloudRuntime
~multiprocess.MultiProcessRuntime

File: docs/source-app/api_reference/storage.rst

#####################
lightning.app.storage
#####################

Lightning Core

.. contents::
:depth: 1
:local:
:backlinks: top

.. currentmodule:: lightning.app.storage

.. autosummary::
:toctree: generated/
:nosignatures:
:template: classtemplate.rst

text
~path.Path
~drive.Drive
~payload.Payload
~mount.Mount

Learn more about Storage

.. raw:: html

.. displayitem::
:header: Learn about the differences between Drive vs Path.
:description: Learn about their differences.
:col_css: col-md-4
:button_link: ../glossary/storage/differences.html
:height: 180
:tag: Basic

.. displayitem::
:header: The Drive Object.
:description: Put, List and Get Files From a Shared Drive Disk.
:col_css: col-md-4
:button_link: ../glossary/storage/drive.html
:height: 180
:tag: Basic

.. displayitem::
:header: The Path Object.
:description: Transfer Files From One Component to Another by Reference.
:col_css: col-md-4
:button_link: ../glossary/storage/path.html
:height: 180
:tag: Intermediate

.. displayitem::
:header: The Mount Object.
:description: Mount an AWS S3 Bucket When Running on the Cloud.
:col_css: col-md-4
:button_link: ../workflows/mount_aws_s3_bucket.html
:height: 180
:tag: Intermediate

.. raw:: html

File: docs/source-app/_templates/classtemplate.rst

.. role:: hidden
:class: hidden-section
.. currentmodule:: {{ module }}

{{ name | underline }}

.. autoclass:: {{ name }}
:members:

File: docs/source-app/_templates/classtemplate_no_index.rst

:orphan:

.. role:: hidden
:class: hidden-section
.. currentmodule:: {{ module }}

{{ name | underline }}

.. autoclass:: {{ name }}
:members:
:noindex:

File: docs/source-app/core_api/lightning_app/communication.rst

:orphan:

##########################################
Communication between Lightning Components
##########################################

Audience: Users that want to create interactive applications.

Level: Intermediate

Prerequisite: Read the :doc:Communication in Lightning Apps article <../../../workflows/access_app_state>.

.. include:: ../../core_api/lightning_app/communication_content.rst

Explore Related Technologies & Topics

View All 9,111+ Repositories →

Discover similar high-velocity repositories, agent skills, and OpenAPI specifications across the ecosystem.

#ai Browse ai catalog → #artificial-intelligence Browse artificial-intelligence catalog → #data-science Browse data-science catalog → #deep-learning Browse deep-learning catalog → #machine-learning Browse machine-learning catalog → #python Browse python catalog → ⚡ MCP Servers Model Context Protocol → 🧠 Agent Skills Prompt & rule files → 🔌 OpenAPI Specs REST & Swagger schemas →