[rllib] Fix LSTM regression on truncated sequences and add regression test #2898
Conversation
This file contains 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
|
Test FAILed. |
|
Test FAILed. |
|
Test PASSed. |
richardliaw
approved these changes
Sep 18, 2018
| @@ -0,0 +1,179 @@ | |||
| """Stateless variant of the CartPole gym environment. | |||
There was a problem hiding this comment.
Stateless..? Might be the wrong word to use here..
|
Test FAILed. |
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Outdated suggestions cannot be applied.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.
What do these changes do?
#2700 regressed LSTM performance by erroneously zeroing out the initial state. This breaks training with truncated sequence lengths (e.g., the tuned pong example).
Incidentally, I also fixed an issue where Pong-ram doesn't pick up the right preprocessor by default.
Testing
Before this fix, the pong a3c example is completely broken. I checked that learning does happen after this fix.
Also added a stateless cartpole env test thanks to @richard4912 , and verified that that test fails before this fix.