You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am curently trying to train UniAD on the Carla dataset.
Due to ressource limitations, I reduced the size of the training dataset to 150 scenes, while keeping the distribution of towns and weather conditions equal.
However, when training stage2 I run into the following problem:
The loss goes up and then drops down again after a while.
This can be seen most clearly in the past_traj loss:
it is however also present in the motion.ade loss as well as others.
Removing the past_traj loss from the training does not solve the issue.
Maybe you can give me an insight in what the problem is here.
Thanks!
The text was updated successfully, but these errors were encountered:
Hello,
first, thanks for your outstanding work!
I am curently trying to train UniAD on the Carla dataset.
Due to ressource limitations, I reduced the size of the training dataset to 150 scenes, while keeping the distribution of towns and weather conditions equal.
However, when training stage2 I run into the following problem:
The loss goes up and then drops down again after a while.
This can be seen most clearly in the past_traj loss:
it is however also present in the motion.ade loss as well as others.
Removing the past_traj loss from the training does not solve the issue.
Maybe you can give me an insight in what the problem is here.
Thanks!
The text was updated successfully, but these errors were encountered: