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 a distributed ML system (ServerlessLLM + RAG).
• Resources: Slides and reading materials related to the course.

---

Last Update

• [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

• Integrate your Task 1 operator into a distributed ML system using ServerlessLLM and RAG.

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.