Torchkit: A toolkit for RL research built using PyTorch

Image credit: Medium

About Torchkit

Torchkit contains a collection (or kit) of tools for Deep RL built with PyTorch. They include

  • an autodifferentiation wrapper around PyTorch for vector valued functions
  • efficient implementations of TRPO and natural policy gradient for high dimensional action spaces
  • running batches of RL experiments
  • creating pretty plots using seaborn
  • some additional RL environments

The Torchkit package source code can be found here.

Licensing Information

© 2018 Carson Eisenach.

Released under the MIT license.

Carson Eisenach
Carson Eisenach
Senior Applied Scientist, Amazon

My research interests include optimization and machine learning.