PyTorch Lightning Documentation

Getting started

• Lightning in 2 steps
• How to organize PyTorch into Lightning
• Rapid prototyping templates

Best practices

• Style guide
• Fast performance tips
• Lightning project template
• Benchmark with vanilla PyTorch

Lightning API

• LightningModule
  • Minimal Example
  • Training
  • Inference
  • LightningModule API
• Trainer
  • Basic use
  • Under the hood
  • Trainer in Python scripts
  • Testing
  • Deployment / prediction
  • Reproducibility
  • Trainer flags
  • Trainer class API

Optional extensions

• Accelerators
  • Implement a custom accelerator
  • Examples
  • Available Accelerators
• Callback
  • Examples
  • Built-in Callbacks
  • Persisting State
  • Best Practices
  • Available Callback hooks
• LightningDataModule
  • Why do I need a DataModule?
  • What is a DataModule
  • LightningDataModule API
  • Using a DataModule
  • Datamodules without Lightning
• Logging
  • Logging from a LightningModule
  • Make a custom logger
  • Control logging frequency
  • Progress Bar
  • Configure console logging
  • Logging hyperparameters
  • Snapshot code
  • Supported Loggers
• Metrics
  • Metrics and devices
  • Implementing a Metric
  • Metric API
  • Class vs Functional Metrics
  • Classification Metrics
  • Regression Metrics
  • NLP
  • Pairwise
• Plugins
  • ApexPlugin
  • NativeAMPPlugin
  • DDPPlugin

Tutorials

• Step-by-step walk-through
• PyTorch Lightning 101 class
• From PyTorch to PyTorch Lightning [Blog]
• From PyTorch to PyTorch Lightning [Video]

API References

• API References
  • Core API
  • Callbacks API
  • Loggers API
  • Profiler API
  • Trainer API
  • Tuner API
  • Utilities API

Bolts

• Bolts

Examples

• Pytorch Ecosystem Examples
• Community Examples
• Autoencoder
• BYOL
• DQN
• GAN
• GPT-2
• Image-GPT
• SimCLR
• VAE

Common Use Cases

• AWS/GCP training
• 16-bit training
• Computing cluster (SLURM)
• Computing cluster
• Child Modules
• Debugging
• Loggers
• Early stopping
• Fast Training
• Hyperparameters
• Learning Rate Finder
• Multi-GPU training
• Multiple Datasets
• Saving and loading weights
• Optimization
• Performance and Bottleneck Profiler
• Single GPU Training
• Sequential Data
• Training Tricks
• Transfer Learning
• TPU support
• Test set
• Inference in Production

Partner Domain Frameworks

• Conversational AI

Community

• Contributor Covenant Code of Conduct
• Contributing
• How to become a core contributor
• PyTorch Lightning Governance | Persons of interest
• Changelog

Indices and tables

• Index

• Search Page