hanveiga
commented
7 years ago Dear Ulrich,
Thank you for the update. I cc'd Daniel in the email just in case. It would be great to have access to the machines at some point - but I understand the cost associated to not running them.
If it we could gain access to the UEdin machines to allocate/deallocate, it would be great. If this is not possible, I can try to set up the environment within the cluster in my university for development/testing purposes, and then once there are things to run, to get an allocation to burn those credits. :-)
I am definitely interested in continuing the project and making a paper submission once appropriate!
Best, Maria
On Sun, Sep 3, 2017 at 9:21 PM, Ulrich Germann notifications@github.com wrote:
@hanveiga https://github.com/hanveiga @da-web https://github.com/da-web
I have Maria's email address but not David's, so I hope by filing this issue I reach both of you.
I deallocated (i.e., shut down but did not remove) the MTM17 Azure instances this morning since they seemed to be idle. Azure charges $0.90 per hour for uptime, irrespective of whether the machine is busy or not, so I decided to shut the machines down for the time being. It is easy to fire them up again if you want to use them, and you can have as many instances as you want (within the limits of the huge allocation of Azure time that UEdin has access to until October 10).
I'll check with Kenneth if there's a way that you can launch your own instances, otherwise I can promise to do it as soon as I get your request and have access to our cluster. I'll be travelling a lot over the next 6 weeks, so there may be periods where I don't have internet access. You can deallocate/shut down machines that you don't need any more by running the command deallocate from within the respective machine.
I'm a bit annoyed at myself that I didn't prepare this project better in the sense of having everything ready to go at the start --- we all would have gotten more out of this project if I had done so --- but to be honest, I was gambling on a stronger involvement of core Marian developers in the project to get things working. This said, it was a pleasure to have you on the project, and I think we've made enough progress during the MT Marathon to continue this. Let's aim for a paper submission to ACL 2018 or another suitable venue.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ugermann/mtm17/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/ABKmYz7dKyiE_d7BLG3etT3YnvB0Zanzks5sevxOgaJpZM4PLRuX .
--
Kind regards, Maria Veiga
PhD Candidate IMATH-ICS University of Zurich url: https://hanveiga.github.io/
@hanveiga @da-web
I have Maria's email address but not David's, so I hope by filing this issue I reach both of you.
I deallocated (i.e., shut down but did not remove) the MTM17 Azure instances this morning since they seemed to be idle. Azure charges $0.90 per hour for uptime, irrespective of whether the machine is busy or not, so I decided to shut the machines down for the time being. It is easy to fire them up again if you want to use them, and you can have as many instances as you want (within the limits of the huge allocation of Azure time that UEdin has access to until October 10).
I'll check with Kenneth if there's a way that you can launch your own instances, otherwise I can promise to do it as soon as I get your request and have access to our cluster. I'll be travelling a lot over the next 6 weeks, so there may be periods where I don't have internet access. You can deallocate/shut down machines that you don't need any more by running the command
deallocatefrom within the respective machine.I'm a bit annoyed at myself that I didn't prepare this project better in the sense of having everything ready to go at the start --- we all would have gotten more out of this project if I had done so --- but to be honest, I was gambling on a stronger involvement of core Marian developers in the project to get things working. This said, it was a pleasure to have you on the project, and I think we've made enough progress during the MT Marathon to continue this. Let's aim for a paper submission to ACL 2018 or another suitable venue.