[RLlib] Reverse learner queue behavior of IMPALA/APPO (consume oldest batches first, instead of newest, BUT drop oldest batches if queue full). #48702
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reverse learner queue behavior of IMPALA/APPO:
consume oldest batches first (consume from left side of deque), instead of newest, BUT drop oldest batches if queue full.
when consuming from the right side (newest batches first) AND the queue size is >1, we run into the danger of consuming "normal" batches most time, but sometimes grabbing a very old batch from the left side of the deque, possibly destabilizing learning suddenly.
We leave the default learner_queue_size at 3 for now.
Why are these changes needed?
Related issue number
Checks
git commit -s
) in this PR.scripts/format.sh
to lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/
under thecorresponding
.rst
file.