Overview

Train, finetune, and pretrain models on Lightning AI. Start with Finetune models for one-click adapters on common base models, or Pretrain models to run from scratch on multi-node clusters.

• Experiment Management — LitLogger, the experiment tracker built into Lightning AI.
• Multi-node — run on more GPUs than fit in a single machine.
• Pretrain models — tools and strategies for large-scale pretraining.
• Finetune models — the ways Lightning AI supports finetuning.
• Using data — read from cloud storage and Lightning datasets during training.
• Examples (full code) — build your own training loop with Fabric.

For the data side, see Data.