AutoML3 :: AutoML for Lifelong Machine Learning Forum

Go back to competition Back to thread list Post in this thread

> I suggest that do not count the number of fail instance.

I am so sorry to say that I often meet some bug and fail on the server. It is not the most serious problem, the most serious problem is that i have success in the docker, so doubt that the server enviroment is different with the docker?

Every team only have two chance to submit each day and total 100 chance. I suggest that do not count the number of fail instance. the number of fail instance can't reflect the true level of the participants. I realy hope you could consider the suggestion.

And I want to know what is the server enviroment, Can you specify a concrete docker?
Sorry for my poor English.

Big thanks to you.

Posted by: Lean-Y @ Aug. 24, 2018, 5:02 a.m.

Thank you for this suggestion

First of all, be sure that the docker used in the server is the same as that specified in the Starting kit, so nothing to worry about this.

Certainly, there are some computing issues with the server that should be due to the heavy load. A solution for this is that whenever a submission fails you can let us know and we will try to re run the submission (we have been doing this with some of the entries). Our apologies for this issues


Posted by: hugo.jair @ Aug. 24, 2018, 7:21 a.m.
Post in this thread