TaoLv commented on issue #17159: Performance regression from 1.4.1 to 1.5.1 URL: https://github.com/apache/incubator-mxnet/issues/17159#issuecomment-568822313 @jonatan1626 Maybe not relate to the regression, but c5.18xlarge should have 2 cpu sockets and 18 cores on each. The env variables should look like: ``` export OMP_NUM_THREADS=36 export KMP_AFFINITY=granularity=fine,noduplicates,compact,1,0 python gluon_resnext50.py --iteration 100000 ``` If you just want to leverage the first cpu socket (like what @wuxun-zhang did on his machine): ``` export OMP_NUM_THREADS=18 export KMP_AFFINITY=granularity=fine,noduplicates,compact,1,0 numactl --physcpubind=0-17 --membind=0 python gluon_resnext50.py --iteration 100000 ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
