NVIDIA Bionemo Inference Runtime

NVIDIA BioNeMo Inference Runtime (BIR) Accelerates biomolecular structure prediction inference for models like OpenFold3 and Boltz-2 on NVIDIA H100 and H200 GPUs. By optimizing discovery-scale workloads, BIR reduces the energy required to generate 31 million protein complex predictions for the AlphaFold Database by an estimated 1.35 GWh - the equivalent of 18,000 EV charges or 5.4 million electric vehicle miles. GitHub - https://nvda.ws/3UKXFy4 Tech blog - https://nvda.ws/4dchK6H