I didn't see any score for out of sample

I select a successful run but i didn’t see any update in out of sample score please can any one clarify

@xgilbert @enzo

It because your In-Sample Runs didn’t managed to run under 20 hours.

If you want we can change a few parameters since its still the first week, but starting the second this will be impossible.

yes please if it possible to change some parameters please I spent a lot of time in this competition. and now what is the quota for each week and total quota for 12 weeks please @enzo

and this will hurt my cumulative score or the first week will be excluded from cumulative score

Which new values do you want?

the new values will depend on rules

1 Your code needs to cover 1-time cross-section in under 16 hours.

2 CPU or GPU ≤ 23 hours runtime to cover the OOS testset (36 dates) with a 5% tolerance

is still first and second applied or only first rule because i didn’t see the second rule except today


please replay as soon as possible and please try hard to rerun my model after modification for week 1 but based on your answer for max quota per week and per all dates i will modify the parameters @enzo

second applies too

And for example if I select train frequency to be 2

It will save the pattern of train or every week it will start training from moon 268

Id say thinking about it in terms of weeks is not correct. Think about it in terms of moon/time id instead. Retrain frequency of 2 means your “train()” function is called every 2 time_ids/moons.

And the retrain frequency is modulo of the moon/id you are predicting on. So if the moon you are predicting on is even, then your train() would be called in that case.

But i’m not part of the team, so that’s just my understanding :slight_smile:

Yes I understand so if we talk about time I’d per moon what should be average execution time to avoid any resources restrictions

@enzo please clarify

Your train function will be called based on the iteration.

If you have a train frequency of 2, then 270, 272, 234…
If you have a train frequency of 3, then 271, 274, 277…
If you have a train frequency of 4, then 272, 276, 280…

And every week we start training from moon 268 correct

No it was just to have a Run on the In-Sample Data that match the Out-of-Sample parameters.

Now the runs will be runs 3 dates at a time, each week.

Okey so can you modify the parameters for me
To be as following

Run #43 CPU , train frequency 4 and it can be run for first week or no

And if no this will affect my cumulative score or no

Could you contact me on Discord please

Please can you send me link o don’t use it before

My user name mohamed_hanafy34