google-deepmind / google-deepmind/rlax
stop_target_gradients default should be True in GAE function
Open
- Dominant language
- Python
- Stars
- 1.4k
- Forks
- 109
- Avg merge
- 2h 15m
- Merged PRs (30d)
- 1
Description
`truncated_generalized_advantage_estimation` should have the `stop_target_gradients` defaulted to `True`
https://github.com/deepmind/rlax/blob/383f93bc8b33c3d1bc28f15e1e07fc5104c790ea/rlax/_src/multistep.py#L279
The `False` case applies only to meta-gradients use case which is rare in the common agents. We should mark this option as defaulted to be `True` to avoid usage bugs.
WDYT?
Contributor guide
Assessment
This issue has not been assessed yet.