Bumping flash attention version to 2.6.3 and adding option for softcap in attention and lm_head logits. #3652
Annotations
2 errors
smoketest (3.10)
Canceling since a higher priority waiting request for 'Smoketest-1374' exists
|
smoketest (3.9)
Canceling since a higher priority waiting request for 'Smoketest-1374' exists
|