Enabling the computation/logging of losses during testing/validation.

allenai / allenact

An open source framework for research in Embodied-AI from AI2.

https://www.allenact.org

Other

320 stars 52 forks source link

Enabling the computation/logging of losses during testing/validation. #362

Closed Lucaweihs closed 1 year ago

Lucaweihs commented 1 year ago

This PR primarily makes changes to the engine.py files to allow for computing losses during validation and testing. This is a feature that is especially useful when one is training with only off-policy datasets. To make this possible I also unified the logging code in runner.py.