marcoabreu commented on issue #11616: Flaky test test_gluon.test_export
URL: 
https://github.com/apache/incubator-mxnet/issues/11616#issuecomment-404237611
 
 
   I tried, what you described was exactly my observation: you run out of cuda 
memory if you run MXNet in parallel. I didn't know we had the memory pool, but 
that certainly explains it. That problem is the reason all our GPU slaves have 
only one executors. Is there any possibility to tweak or disable that pool?
   
   We're not in state where we can give actual guarantees at all, that's why I 
had it in quotes :)
   
   No, running the tests in parallel (which is only the case on windows CPU) 
only reveals the problems we're having if we actually don't run in sequence. 

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

Reply via email to