Industry
Train Qwen3 on AWS Trainium with torchtitan
Train Qwen3 on AWS Trainium with torchtitan three ways on a single trn2.48xlarge: the Hugging Face Transformers v5 modeling backend, torchtitan's native Qwen3, and NKI mixture-of-experts kernels. The
Train Qwen3 on AWS Trainium with torchtitan three ways on a single trn2.48xlarge: the Hugging Face Transformers v5 modeling backend, torchtitan's native Qwen3, and NKI mixture-of-experts kernels. The same torchrun launcher and config structure work unchanged across all three paths, including loading and saving published Hugging Face safetensors.
Related
- Mixture-of-Experts kernel engineering on Amazon SageMaker AI
- Deploying Kimi K3 on Amazon SageMaker HyperPod and Amazon EKS
- Deploying Kimi K3 on AWS
- How Deepgram optimized voice AI at scale with AWS Trainium
Source: AWS ML Blog | 2026-10-21