Evaluation Puck Velocities

[Thifhi]

Hi,

We have realized that the puck velocities in the datasets from remote evaluations are about ~20 times larger than what we get in the local evaluations. For instance, a random hit episode has an observed puck velocity in the range of [0, 0.5] in local evaluations, while it is in the range of ~[0, 10] in the server evalutions. Since the range difference seems too large to be explained with noise, is this maybe a bug in the evaluation?

[cube1324]

Hello,
You are indeed correct, there was a bug in the computation of the velocity. Thank you for spotting it, you saved us a lot of time.
We fixed the bug for the evaluation on the server. I will rerun the so far submitted images later today, so you can work with the proper data.
Sorry for the inconvenience, the runs should be done tomorrow.