.svg)
Amazon SageMaker AI announces training session APIs for serverless model customization
Amazon SageMaker AI Serverless Model Customization
Amazon SageMaker AI now supports training session APIs, allowing full control over post-training algorithms without managing GPU infrastructure. Managed techniques like SFT, DPO, and reinforcement learning methods are available, enabling custom training loops and operations like forward_backward(), optim_step(), and sample() as API calls. Billing is per token processed, excluding local computation like reward scoring.
What to do
- Install the SageMaker Python SDK
- Clone the SageMaker training sessions cookbook
- Launch a notebook in SageMaker Studio and connect a coding agent
Source: AWS release notes
If you need further guidance on AWS, our experts are available at AWS@westloop.io. You may also reach us by submitting the Contact Us form.
