Skip to content

Latest commit

 

History

History
47 lines (31 loc) · 2.33 KB

README.md

File metadata and controls

47 lines (31 loc) · 2.33 KB

Machine Learning Systems - 2024/2025

Welcome to the repository for the Machine Learning Systems course (INFR11269) for the 2024/2025 academic year. This course focuses on building and deploying machine learning systems, with hands-on programming tasks, paper writing, and peer reviews.

The full course schedule, assessments, and additional details are available in the official course page:

Machine Learning Systems - 2024/2025

Repository Structure

  • Task-1: Part 1, Implementing machine learning operators with GPU programming.
  • Task-2: Part 2, Integrating the operator into an end-to-end RAG serving pipeline.
  • Resources: Slides and reading materials related to the course.

Last Update

  • [07/03/2025] We have released task 2. Please check the live recording of today's lecture for details. (The slides will be uploaded later)
  • [11/02/2025] Update instructions of pytorch demo. If you encounter a No disk space error, try logging into Interactive mode first and installing the environment on the node.
  • [05/02/2025] We have uploaded the code template for the first part of the assessment into the task-1 folder. Additionally, we have relocated the pytorch-demo to the resources directory and have included materials for gpu-programming in the same directory. The part 1 specification in under the Assessment section on Learn.

Course Tasks

Task 1: Triton/Cupy Operator

  • Implement an ML operator using Triton/Cupy.
  • Learn about performance optimization and profiling.

Task 2: Integration into Distributed System

  • Deploy an end-to-end RAG serving system.
  • Measure the system performance.
  • Optimize the deployment with a request queue and batcher.

Paper Writing

  • Write a paper documenting your work on both tasks in the format of a NeurIPS or ICML paper.

Course Schedule

The course consists of 10 weeks of lectures and Q&A sessions. Each week has the following structure:

  • Lectures: Core topics presented by the primary and guest lecturers.
  • Q&A Sessions: Focused on solving problems, demos, and discussing task-related questions.

The full course schedule is available here.