1 paper · 1 filter
Pranshu Chaturvedi, Parth Shroff, Tarun Suresh +2
High-throughput long-context generation is one of the central challenges for large language models. Generation is typically memory-bandwidth-bound rather than compute-bound: each d…