AnswerDotAI/cold-compress

Record Model Speed in evals

griff4692 opened this issue · 0 comments

This should be end-to-end (not just tokens / second during generation).