Cleaner PyTorch training code that scales to more GPUs.
Guides the agent in organizing PyTorch training code into Lightning modules with managed training, validation, and distributed execution.
When to use it
Use when developing organized training loops with built-in callbacks and logging, or switching between single-GPU, multi-GPU, and TPU training.
It changes how the agent organizes PyTorch training code, using Lightning modules and Trainer-based workflows.
What you provide
Uses Python to run the training script; no Python version requirement is given.
The training examples use the lightning package, and the quick start gives its installation command.
The core examples import torch for neural-network layers, data loading, and optimization.
This skill
lightning
Installs the lightning package