Refactor multistep return estimators to use scan instead of a for loop. When

google-deepmind / rlax

https://rlax.readthedocs.io

Apache License 2.0

1.24k stars 85 forks source link

Refactor multistep return estimators to use scan instead of a for loop. When #101

Closed copybara-service[bot] closed 2 years ago

copybara-service[bot] commented 2 years ago

Refactor multistep return estimators to use scan instead of a for loop. When training on long sequences this can dramatically reduce compilation time.