Building Language Models on AWS

Aperçu

Amazon SageMaker empowers data scientists to efficiently prepare, build, train, deploy, and monitor machine learning models. It offers a robust suite of capabilities, including the utilization of distributed training libraries, open-source models, and foundational models.

Particularly, this advanced 5.5-hour course is designed for experienced data scientists keen on mastering the construction of language models using large text corpora. The course covers storage, ingestion, and training alternatives, deployment challenges, and methodologies for customizing foundational models for generative AI tasks with Amazon SageMaker Jumpstart.

The course delivers practical insights through textual instructions, illustrative graphics, knowledge checks, and hands-on video demonstrations that you can replicate within your AWS account.

You will learn and apply best practices for data storage and ingestion, explore both data and model parallelism, discuss performance improvement options like the Amazon SageMaker Training Compiler and Elastic Fabric Adapter, and delve into optimization techniques for large language models.

• Apply best practices for managing extensive text data for distributed training.
• Understand and implement data and model parallelism to enhance training efficiency.
• Explore options on Amazon SageMaker to boost training performance.
• Optimize and deploy large language models effectively.
• Customize foundational models on SageMaker Jumpstart for specific AI tasks.

This course is most beneficial for data scientists and machine learning engineers. Attendees are expected to have over a year of experience in natural language processing and model training and tuning.

They should also have an intermediate understanding of Python programming, and familiarity with AWS Technical Essentials and Amazon SageMaker Studio.

By the conclusion of this course, participants will possess the skills to build, train, and fine-tune high-performance language models on AWS using SageMaker, setting a strong foundation for future learning and application in real-world scenarios.

Matières