Benchmarking LLM Fine-Tuning on Different HPC Systems

April 7, 2026

I guess I need to start somewhere, and I think the post below — which I worked on with Flavio Hafner and others — is a good opening for this blog. Last year, we benchmarked LLM fine-tuning performance across Dutch and Princeton HPC systems, and we’ve been hearing from more groups interested in running the benchmark on their clusters.

Fine-tuning infrastructure varies quite a bit, and it’s helpful to have a reliable conversation (and collaboration!) starter.

Read the full post on the eScience Center blog

View the benchmark repo on GitHub