22 KiB
The Deep Learning framework to train, deploy, and ship AI products Lightning fast.
NEW- Lightning 2.0 is featuring a clean and stable API!!
Lightning.ai • PyTorch Lightning • Fabric • Lightning Apps • Docs • Community • Contribute •
Train and deploy with PyTorch Lightning
PyTorch Lightning is just organized PyTorch- Lightning disentangles PyTorch code to decouple the science from the engineering.
How to use PyTorch Lightning
Step 1: Add these imports
import lightning as L
import os
import torch
from torch import nn
import torch.nn.functional as F
from torchvision.datasets import MNIST
from torch.utils.data import DataLoader, random_split
from torchvision import transforms
Step 2: Define a LightningModule (nn.Module subclass)
A LightningModule defines a full system (ie: a GAN, autoencoder, BERT or a 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, y = 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
Note: Training_step defines the training loop. Forward defines how the LightningModule behaves during inference/prediction.
Step 3: Train!
dataset = MNIST(os.getcwd(), download=True, transform=transforms.ToTensor())
train, val = random_split(dataset, [55000, 5000])
autoencoder = LitAutoEncoder()
trainer = L.Trainer()
trainer.fit(autoencoder, DataLoader(train), DataLoader(val))
Advanced features
Lightning has over 40+ advanced features designed for professional AI research at scale.
Here are some examples:
Highlighted feature code snippets
# 8 GPUs
# no code changes needed
trainer = Trainer(max_epochs=1, accelerator="gpu", devices=8)
# 256 GPUs
trainer = Trainer(max_epochs=1, accelerator="gpu", devices=8, num_nodes=32)
Train on TPUs without code changes
# no code changes needed
trainer = Trainer(accelerator="tpu", devices=8)
16-bit precision
# no code changes needed
trainer = Trainer(precision=16)
Experiment managers
from lightning import loggers
# 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())
# neptune
trainer = Trainer(logger=loggers.NeptuneLogger())
# ... and dozens more
EarlyStopping
es = EarlyStopping(monitor="val_loss")
trainer = Trainer(callbacks=[es])
Checkpointing
checkpointing = ModelCheckpoint(monitor="val_loss")
trainer = Trainer(callbacks=[checkpointing])
Export to torchscript (JIT) (production use)
# torchscript
autoencoder = LitAutoEncoder()
torch.jit.save(autoencoder.to_torchscript(), "model.pt")
Export to ONNX (production use)
# 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)
Pro-level control of optimization (advanced users)
For complex/professional level work, you have optional full control of the optimizers.
class LitAutoEncoder(L.LightningModule):
def __init__(self):
super().__init__()
self.automatic_optimization = False
def training_step(self, batch, batch_idx):
# access your optimizers with use_pl_optimizer=False. Default is True
opt_a, opt_b = self.optimizers(use_pl_optimizer=True)
loss_a = ...
self.manual_backward(loss_a, opt_a)
opt_a.step()
opt_a.zero_grad()
loss_b = ...
self.manual_backward(loss_b, opt_b, retain_graph=True)
self.manual_backward(loss_b, opt_b)
opt_b.step()
opt_b.zero_grad()
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).
Examples
Self-supervised Learning
Convolutional Architectures
Reinforcement Learning
GANs
Classic ML
Read more about PyTorch Lightning
Scale PyTorch With Lightning Fabric
Fabric allows you to scale any PyTorch model to distributed machines while maintianing full control over your training loop. Just add a few lines of code and run on any device! Use this library for complex tasks like reinforcement learning, active learning, and transformers without losing control over your training code.
Learn more about Fabric
With just a few code changes, run any PyTorch model on any distributed hardware, no boilerplate!
- Easily switch from running on CPU to GPU (Apple Silicon, CUDA, …), TPU, multi-GPU or even multi-node training
- Use state-of-the-art distributed training strategies (DDP, FSDP, DeepSpeed) and mixed precision out of the box
- All the device logic boilerplate is handled for you
- Designed with multi-billion parameter models in mind
- Build your own custom Trainer using Fabric primitives for training checkpointing, logging, and more
+ import lightning as L
import torch
import torch.nn as nn
from torch.utils.data import DataLoader, Dataset
class PyTorchModel(nn.Module):
...
class PyTorchDataset(Dataset):
...
+ fabric = L.Fabric(accelerator="cuda", devices=8, strategy="ddp")
+ fabric.launch()
- device = "cuda" if torch.cuda.is_available() else "cpu
model = PyTorchModel(...)
optimizer = torch.optim.SGD(model.parameters())
+ model, optimizer = fabric.setup(model, optimizer)
dataloader = DataLoader(PyTorchDataset(...), ...)
+ dataloader = fabric.setup_dataloaders(dataloader)
model.train()
for epoch in range(num_epochs):
for batch in dataloader:
input, target = batch
- input, target = input.to(device), target.to(device)
optimizer.zero_grad()
output = model(input)
loss = loss_fn(output, target)
- loss.backward()
+ fabric.backward(loss)
optimizer.step()
lr_scheduler.step()
Read more about Fabric
Build AI products with Lightning Apps
Once you're done building models, publish a paper demo or build a full production end-to-end ML system with Lightning Apps. Lightning Apps remove the cloud infrastructure boilerplate so you can focus on solving the research or business problems. Lightning Apps can run on the Lightning Cloud, your own cluster or a private cloud.
Browse available Lightning apps here
Learn more about apps
Build machine learning components that can plug into existing ML workflows. A Lightning component organizes arbitrary code to run on the cloud, manage its own infrastructure, cloud costs, networking, and more. Focus on component logic and not engineering.
Use components on their own, or compose them into full-stack AI apps with our next-generation Lightning orchestrator. to package your code into Lightning components which can plug into your existing ML workflows.
Run your first Lightning App
-
Install a simple training and deployment app by typing:
# install lightning pip install lightning lightning install app lightning/quick-start -
If everything was successful, move into the new directory:
cd lightning-quick-start -
Run the app locally
lightning run app app.py -
Alternatively, run it on the public Lightning Cloud to share your app!
lightning run app app.py --cloud
Apps run the same on the cloud and locally on your choice of hardware.
run the app on the --cloud
lightning run app app.py --setup --cloud
Learn more about Lightning Apps
Continuous Integration
Lightning is rigorously tested across multiple CPUs, GPUs, TPUs, IPUs, and HPUs and against major Python and PyTorch versions.
*Codecov is > 90%+ but build delays may show less
Current build statuses
| System / PyTorch ver. | 1.11 | 1.12 | 1.13 | 2.0 |
|---|---|---|---|---|
| Linux py3.9 [GPUs] | - | Soon | ||
| Linux py3.9 [TPUs] | - |  |
Soon | |
| Linux py3.8 [IPUs] | - | - | Soon | |
| Linux py3.8 [HPUs] | - | - | Soon | |
| Linux (multiple Python versions) |  |
|
|
Soon |
| OSX (multiple Python versions) | |
|
|
Soon |
| Windows (multiple Python versions) | |
|
|
Soon |
Install
Simple installation from PyPI
pip install lightning
Other installation options
Install with optional dependencies
pip install lightning['extra']
Conda
conda install lightning -c conda-forge
Install stable version
Install future release from the source
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)
pip install https://github.com/Lightning-AI/lightning/archive/refs/heads/master.zip -U
or from testing PyPI
pip install -iU https://test.pypi.org/simple/ pytorch-lightning
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.
- 590+ active 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: