Build efficient agents with knowledge distillation on Amazon SageMaker AI
According to AWS. Learn how to use sequence-level knowledge distillation to fine-tune a shopping agent on Amazon SageMaker AI.
- • Use sequence-level knowledge distillation to build efficient agents.
- • Generate multi-step trajectories with a frontier teacher model.
- • Fine-tune a smaller student model using serverless training on Amazon SageMaker AI.
Sequence-Level Knowledge Distillation
Sequence-level knowledge distillation uses a teacher model to generate high-quality responses, then trains a student model on those responses with supervised fine-tuning. It needs only the teacher's text with no access to teacher logits or reinforcement-learning infrastructure, making it work even with a closed or API-only teacher.
Solution Overview and Pipeline
A frontier teacher on Amazon Bedrock acts as the ShoppingBench agent to generate trajectories, and a benchmark scorer keeps only the ones that solve the task. Those solved trajectories fine-tune a student model with serverless supervised fine-tuning on Amazon SageMaker AI model customization.
- 1. aws.amazon.comOpen source