R-Ceph / pearl_reproduce

Meta RL codebase for Unstable Baselines

Meta Reinforcement Learning Re-implementations for Unstable Baselines

Stable Algorithms (Runnable and has equivalent performance to that of the original implementations)

• Efficient Off-policy Meta-learning via Probabilistic Context Variables (PEARL)

Quick Start

• Requires the Unstable Baselines parent module to run. See Unstable Baselines for more details.

Results

• The Tensorboard logs of some experimental runs has been included in the results directory.
• Test Machine Information
  • Ubuntu 20.04
  • Intel I9-12900KF
  • NVIDIA RTX 3080
  • Crucial DDR5 96 GigaBytes