---
format: "aidr-story-markdown/v1"
id: "b5c718abbbf002f946f66d6730314c11844bb8ff750a6662307d3c2446d87dd8"
canonical_url: "https://aidr.today/b5c718ab?lang=en"
title: "Train Qwen3 on AWS Trainium with torchtitan"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-21T20:00:22.000Z"
category: "Frameworks"
topics: ["llm","framework"]
source_urls: ["https://aws.amazon.com/blogs/machine-learning/train-qwen3-on-aws-trainium-with-torchtitan/"]
summary: "Qwen3 can be trained on AWS Trainium using torchtitan through three methods on a single trn2.48xlarge: the Hugging Face Transformers v5 modeling backend, torchtitan's native Qwen3, and NKI mixture-of-experts kernels, with a consistent torchrun launcher and config structure across all paths."
---

# Train Qwen3 on AWS Trainium with torchtitan

> [Open the canonical story](<https://aidr.today/b5c718ab?lang=en>)

**Published:** 2026-10-21T20:00:22.000Z
**Category:** Frameworks
**Topics:** llm, framework

## Summary

Qwen3 can be trained on AWS Trainium using torchtitan through three methods on a single trn2\.48xlarge: the Hugging Face Transformers v5 modeling backend, torchtitan's native Qwen3, and NKI mixture\-of\-experts kernels, with a consistent torchrun launcher and config structure across all paths\.

## Sources

- [Story source](<https://aws.amazon.com/blogs/machine-learning/train-qwen3-on-aws-trainium-with-torchtitan/>)