[rllib] Fix torch c51 dqn #16716
Merged
[rllib] Fix torch c51 dqn #16716
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
|
Hmm @Souphis it looks like Is it possible to investigate that failure? |
|
I don't think that this PR is related to this failure, but I will investigate this. |
|
@Souphis , thanks for this PR! Taking a look at the failing test. I think it's not related to this PR, though, so should be good, but I'll confirm. ... |
|
@sven1977 I ran tests once again locally, both td3 and dqn passed them. So, I think that this error is not related to this PR. |
|
Merged with master, which has a fix for this test case. Waiting for tests to pass again. ... |
|
@Souphis , agree. Will merge as soon as everything passes. Should be today :) |
sven1977
approved these changes
Jul 13, 2021
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Outdated suggestions cannot be applied.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.
Why are these changes needed?
Currently, it is not possible to train a c51 torch agent. The output shape of the value head in DqnTorchModel should be equal to the num_atoms (now it is equal to 1). Also in the current implementation, the noisy option doesn't affect the value head. There are also minor fixes in QLoss, like target probs should be detached before loss calculation.
Related issue number
Checks
scripts/format.shto lint the changes in this PR.