{"author":"chenyu","author_email":"chenyu@fastmail.com","author_time":1715296532,"commit_time":1715296532,"committer":"GitHub","committer_email":"noreply@github.com","hash":"5de4a46f10405c84bcb893e21770bcc722a5a953","message":"re-enable gpt2 half/beam mac benchmark (#4496)\n\n* re-enable gpt2 half/beam mac benchmark\r\n\r\nfrom fuzzer it seems to be flaky due to numerical issue, not kernel bug. we used to have half in splitted reduce.\r\n\r\nrun this in M1 Max for 20 loops and it's fine\r\n\r\n* that should be jitted","parents":["a2e2ba380c6d476e7368dfd7c22cba2cf88b66b1"],"tree_hash":"d1f54c692a9afbefa9f5c3db2ed6d87491d59004"}