Hello @thomas.yu and @brucehoff Because of disk latency (see https://www.synapse.org/#!Synapse:syn4224222/discussion/threadId=1417) my model apparently does not have enough time to finish all the iterations. However, according to the logs it already significantly improves my results. Unfortunately (my bad) because of disk space limitations I copy the best model from /scratch to /modelState after all the iterations are done (cp /scratch/best_model* /modelState). So, if I cancel now all the docker container via web interface I will loose the best model. Could you please just kill the Python training script and make sure that the best model was copied to /modelState (next line in train.sh) so I can download it and use for scoring? Submission ID is 7894930. Thank you so much! Best, Yaroslav

Created by Yaroslav Nikulin (Therapixel) ynikulin
> When exactly is the official deadline, which timezone? Round 1 ends midnight tonight, Pacific Standard Time. That is the time at which no new submissions can be created in Round 1. That is *not* necessarily the time at which running submissions will be terminated, a process dictated by our need to reclaim the servers for maintenance for Round 2. So I recommend you simply let your training submission continue running.
Ok, I see. I hope it will finish in 4-5 hours. When exactly is the official deadline, which timezone?
I'm sorry but we don't have the ability to modify your submitted, running model. Why not wait and see if your running submission finishes up by the end of round 1? If not, resubmit on the first day of round 2, which should give plenty of time to regenerate and collect your results.

Custom cancel request - easy to do, please help page is loading…