Closed cuixiaom closed 6 years ago
Thanks for your pull request. It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).
:memo: Please visit https://cla.developers.google.com/ to sign.
Once you've signed, please reply here (e.g. I signed it!
) and we'll verify. Thanks.
I signed it!
We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for the commit author(s). If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google. In order to pass this check, please resolve this problem and have the pull request author add another comment and the bot will run again.
@cuixiaom I think you need to associate the commit with the email address you signed the CLA.
Do you know what value user should set in order to get a good performance on CPU? some comparison among different configurations would be very useful.
Thanks, Rui
@oahziur I have CLA covered with this email address. Please verify. For your question, we did see the performance variance with different CPU threads. For my experience, the threads should be closed to the number of cores of CPU. but sometimes, it will not true depends on the availability of resources. it depends on users' the system configuration.
@cuixiaom I think the problem with CLA is that you need to associate the email address with your git commit, otherwise, I think the CLA check won't pass. For example, if I open your commit, it shows unrecognized author.
@oahziur I have create a new PR since this PR is messed up with CLA and account information issue. Please check the new PR I created. https://github.com/tensorflow/nmt/pull/194 Thanks!
…threads to control threads of session
On CPU side of training, the config_proto.inter_op_parallelism_threads and config_proto.intra_op_parallelism_threads significantly impact performance, the default value of these two parameter doesn't guarantee the best performance, it should have capability to let user to decide these two parameters.