Benchmarking LLM finetuning and multi-node NCCL communication

Benchmarks for finetuning LLMs on HPC systems and investigating performance bottlenecks.

August 2025 · Flavio Hafner, Mattie Niznik