{"author":"chenyu","author_email":"chenyu@fastmail.com","author_time":1704997266,"commit_time":1704997266,"committer":"GitHub","committer_email":"noreply@github.com","hash":"93e3f952aa20370320b35a1a21f608fc72baebdf","message":"use BEAM=2 instead of BEAM=4 in cuda ci gpt2 (#3089)\n\nBEAM=2 is faster and less search time. investigating why BEAM2+BEAM4 is slower than BEAM2 alone","parents":["f502c9b08f8dd9bbbe3ab39c51e5c19bbce2ad7e"],"tree_hash":"c1ad50cdff4e76c7fab89b10ed7f07e64890c92c"}