Exploring GPU-Accelerated Search for Tens of Billions of Proteins in MMseqs2 #1070
Description
Dear Professors Steinegger @martin-steinegger and Mirdita @milot-mirdita,
It is exciting to hear about the advancements in GPU-accelerated search in MMseqs2! Given the scenario of working with tens of billions of proteins for self-alignment, is GPU-accelerated search currently suitable for such tasks? Additionally, could you provide the recommended memory and GPU configurations for optimal performance? Have you tested this approach with datasets of tens of billions of proteins? Furthermore, what sensitivity level is used for GPU-accelerated searches, and is it possible to adjust this parameter?