rahul003 edited a comment on issue #11777: Distributed Training: looks like async from the log although setting the kv-store=dist_device_sync URL: https://github.com/apache/incubator-mxnet/issues/11777#issuecomment-405760529 That order doesn't look normal in my experience. What network are you training? Also regarding the last validation part, do you see that the training hangs at that point on the last epoch? That could explain why some processes go ahead and finish validation while others are stuck because they have more batches than other worker machines (sync mode). If so, can you ensure you have the changes from this commit in your code. https://github.com/apache/incubator-mxnet/pull/10435/
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services
