search

shmsw25 / bart-closed-book-qa

A BART version of an open-domain QA model in a closed-book setup
120 stars 14 forks source link

RuntimeError: NCCL Error 2: unhandled system error #8

Closed zhenyuhe00 closed 3 years ago

zhenyuhe00 commented 3 years ago

Hi, Thank you for your nice code. I ran your code with 8 Nvidia 2080ti GPUs and but this error occurred. Do you know anything about it? Thanks in advance!

shmsw25 commented 3 years ago

Hi @hzy-hzy, sorry I am not aware of this error - the error seems to be a problem of pytorch and hardward mismatch, rather than this specific code. But looks like there're related discussions, including this pytorch issue, which would help!

zhenyuhe00 commented 3 years ago

That's very kind of you. Thanks!