Fast ML Inference

Inference is a crucial step in the machine learning workflow—especially in High-Energy Physics—where data is produced at extremely high rates and needs to be processed rapidly and efficiently. As part of the NGT, we’ve been exploring methods for fast ML inference optimized for heterogeneous architectures.

During the hackathon, we made significant progress toward our objectives. Our focus was on extending the capabilities of SOFIE—a tool developed at CERN for fast ML inference. We contributed by developing new features, addressing user requests, fixing bugs, and improving overall usability.

Highlights and Progress from the Hackathon: