pytorch-lightning

SUMMARY

What it does

This skill provides procedural knowledge for using PyTorch Lightning to structure and scale deep learning training. It covers organizing PyTorch code into LightningModules, configuring Trainers for multi-GPU/TPU training, implementing data pipelines with LightningDataModules, using callbacks, integrating experiment loggers (TensorBoard, W&B, MLflow, Comet, CSV), and applying distributed training strategies (DDP, FSDP, DeepSpeed). The skill includes templates and reference documentation for common patterns and best practices. It is intended for software development, testing, data analysis, scientific research, document processing, writing, DevOps, and agent orchestration tasks.