This document is relevant for: Trn2, Trn3

TorchTitan training tutorials#

End-to-end walk-throughs for common TorchTitan pre-training workloads on AWS Trainium.

Basic training tutorial

A first, complete TorchTitan training run on Trainium.

SimpleFSDP training with Llama and DeepSeek tutorial

Compiler-based sharding for two real model families.

GPT-OSS training tutorial

End-to-end GPT-OSS training with TorchTitan on Trainium.

This document is relevant for: Trn2, Trn3