Added PPO+LSTM, plus training example#39
Open
roger-creus wants to merge 5 commits intoRLE-Foundation:mainfrom
Open
Added PPO+LSTM, plus training example#39roger-creus wants to merge 5 commits intoRLE-Foundation:mainfrom
roger-creus wants to merge 5 commits intoRLE-Foundation:mainfrom
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
I have added a new agent -- PPO + LSTM, together with the new EpisodicRolloutBuffer, which is similar to VanillaRolloutBuffer but samples entire trajectories instead of random transitions in order to train the LSTM appropriately.
I have also added an example notebook to train it on Atari - Space Invaders, which achieves the following results:

In this case, it performs very similarly to vanilla PPO:
Motivation and Context
PPO LSTM can achieve better performance than PPO in partially observed environments.
Types of changes
Checklist
make format(required)make check-codestyleandmake lint(required)make pytestandmake typeboth pass. (required)make doc(required)Note: You can run most of the checks using
make commit-checks.Note: we are using a maximum length of 127 characters per line