HLS4ML Tutorial

Friday May 29, 2026, 2:00 PM → 6:00 PM Europe/Zurich

40/4-C01 (CERN)

This tutorial provides a practical introduction to hls4ml for FPGA-based machine learning inference. Participants will go through the main hls4ml workflow using small neural network examples, learn how key configuration choices affect latency and resource usage, and explore basic optimization techniques such as quantization and pruning.

The hands-on parts of the tutorial will run on a CERN-curated Jupyter notebook platform, provided thanks to the NGT Task 1.1 team. Registered participants will be able to access the notebooks using their CERN guest accounts.

The platform is here and the instructions for it are at the end of the slides. 

You will need to copy a setup script from the github repo. 

The session is designed as a practical first workflow rather than a complete FPGA architecture or hls4ml developer-level course.

FDF-hls4ml .pdf

2:00 PM → 2:35 PM hls4ml: Introduction

2:35 PM → 2:40 PM hls4ml: Tutorial setup

2:40 PM → 3:15 PM tutorial: Part 1: MLP workflow

3:15 PM → 3:30 PM Break

3:30 PM → 4:30 PM tutorial: Part 2: CNN workflow

4:30 PM → 4:40 PM tutorial: Part 3: Deployment demo

4:40 PM → 5:00 PM hls4ml: Q&A