Eclectic-Sheep / sheeprl

Distributed Reinforcement Learning accelerated by Lightning Fabric
https://eclecticsheep.ai
Apache License 2.0
303 stars 31 forks source link

Feature/replay ratio #247

Closed belerico closed 6 months ago

belerico commented 6 months ago

Summary

This PR introduces the replay_ratio as specified in #223. In particular we have removed both the per_rank_gradient_steps and train_every, since those two can be computed with the single output from the newly introduced Ratio class. We have also modified how the Dreamer-V1 family algos compute the exploration actions, while removing that from Dreamer-V2 and Dreamer-V3 since those are unused by those algos.

Type of Change

Please select the one relevant option below:

Checklist

Please confirm that the following tasks have been completed:

Screenshots or Visuals (Optional)

If applicable, please provide screenshots, diagrams, graphs, or videos of the changes, features or the error.

Additional Information (Optional)

Please provide any additional information that may be useful for the reviewer, such as:

Thank you for your contribution! Once you have filled out this template, please ensure that you have assigned the appropriate reviewers and that all tests have passed.