Amazon Web Services (AWS Machine Learning Blog) · Serverless FFT

Full Fine-Tuning on SageMaker AI Serverless Model Customization

AWS has added Full Fine-Tuning (FFT) to Amazon SageMaker AI Serverless Model Customization, so users can run full-parameter fine-tuning without provisioning GPUs or configuring distributed training. The post compares FFT with LoRA by fine-tuning Qwen3-4B on the 110K-pair Magicoder-Evol-Instruct dataset, reporting that FFT reached 90.0 percent on LLM-as-Judge semantic correctness for code generation at 100K+ examples. It also includes SDK code showing that switching between FFT and LoRA changes only the training type setting.

Open the source ↗

Similar Developer Tools releases

More from Amazon Web Services (AWS Machine Learning Blog)

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.