Full Fine-Tuning (FFT) on SageMaker AI Serverless Model Customization
TL;DR
Amazon SageMaker AI Serverless Model Customization now supports Full Fine-Tuning (FFT). This post explores when FFT outperforms LoRA across key benchmarks, examines the spectral properties of model weight matrices behind the quality gap, shares practical hyperparameter tuning guidelines, and shows how to get started.
Nauti's Take
For a small team, the first test should be a controlled A/B comparison using the same dataset, budget, and evaluation set. FFT may improve results on demanding customization tasks, but it also requires more compute and tuning.
AWS benchmarks provide a useful starting point, not reliable evidence for your domain.