[bug-fix] Fix issue where NaNs are outputted by the policy when training Match3 #4664
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Outdated suggestions cannot be applied.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.
Proposed change(s)
In certain environments (particularly with large action spaces), discrete actions can become
NaNoccasionally. This is especially evident with the Match3 environment.This PR adds a small epsilon value to any division or log function that may converge very close to zero, particularly in the
MultiCategoricalDistributionandCategoricalDistInstance.Types of change(s)
Checklist
Other comments