PyTorch Lightning is an open-source library that organises PyTorch training code, handling the repetitive engineering so the same model can scale from a laptop to many GPUs.
In plain words
Plain PyTorch leaves you to write training loops, mixed precision, multi-GPU and distributed training yourself, often again for each project. Lightning splits the work: you put your model and training step in a LightningModule, and a Trainer runs it, taking care of the hardware. Features are switched on with Trainer settings, and experts can still take full control of the loop when they need to.
A quick example
Imagine a student who trained a model on one GPU and now has access to eight. With Lightning, the model code stays the same; they change the Trainer's settings to use more devices, and it handles splitting the work.
Where you'll meet it
- Research code that must scale from one GPU to many.
- Fine-tuning and pretraining projects built on PyTorch.
Related
Sources
- DocsLightning-AI/pytorch-lightning, Lightning AI (GitHub) · read 29 Sept 2026
- DocsLightning in 15 minutes, PyTorch Lightning documentation source · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.