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
There are currently quite a few deep copies in the code base that were added when debugging the initial code implementations in quite a conservation way.
In eaa9f40, I've removed those that are definitely safe to remove. All tests still pass.
Some deepcopies are likely best remaining, such as those used in callbacks.
Whether others can be removed largely comes down to how much in-place operations we want to allow. Whatever we do, I think we need to decide on a consistent rule.
The main thing we need to watch out for is particles.filtered pointing to particles.proposed and then making in-place changes to filtered which will change the proposed particles that still need to be kept around for the call back.
An alternative way around this would be have to have multiple callback opportunities, one after predict one after update so you know you have the correct distribution at that time (this would involve removing intermediate)
The text was updated successfully, but these errors were encountered:
There are currently quite a few deep copies in the code base that were added when debugging the initial code implementations in quite a conservation way.
In eaa9f40, I've removed those that are definitely safe to remove. All tests still pass.
Some deepcopies are likely best remaining, such as those used in callbacks.
Whether others can be removed largely comes down to how much in-place operations we want to allow. Whatever we do, I think we need to decide on a consistent rule.
The main thing we need to watch out for is
particles.filtered
pointing toparticles.proposed
and then making in-place changes tofiltered
which will change theproposed
particles that still need to be kept around for the call back.An alternative way around this would be have to have multiple callback opportunities, one after
predict
one afterupdate
so you know you have the correct distribution at that time (this would involve removingintermediate
)The text was updated successfully, but these errors were encountered: