intermediate/rpc_tutorial: fix REINFORCE return calculation across observers#3893
Open
getchebarne wants to merge 1 commit into
Open
intermediate/rpc_tutorial: fix REINFORCE return calculation across observers#3893getchebarne wants to merge 1 commit into
getchebarne wants to merge 1 commit into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/tutorials/3893
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
7 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
The
finish_episodecode block inintermediate/rpc_tutorialmerges every observer's rewards into one flat list before doing the reverse discounted sum. BecauseRnever gets reset between observers, the early steps of observer 1's trajectory end up with observer 2's whole return tacked onto their tail. Those rewards came from an independent environment, which is conceptually wrong in many levels.The fix: compute the discounted return per observer, then concatenate. Everything around it (running reward, buffer clear, normalization, loss, optimizer step) stays the same. Single-observer behavior is unchanged.
Same change is going up against the mirrored runnable code in pytorch/examples: http://github.com/pytorch/examples/pull/1422.
Checklist