Wrapper around dm_control to provide a gym like interface and vice-versa
-
Updated
Jul 31, 2018 - Python
Wrapper around dm_control to provide a gym like interface and vice-versa
Convert DeepMind Control Suite to OpenAI gym environments.
PyTorch implementation of Soft Actor-Critic + Autoencoder(SAC+AE)
Training code and evaluation benchmarks for the "Self-Supervised Policy Adaptation during Deployment" paper.
SUNRISE: A Simple Unified Framework for Ensemble Learning in Deep Reinforcement Learning
RAD: Reinforcement Learning with Augmented Data
A Deep Reinforcement Learning (DeepRL) package for RL algorithm developers.
Implementation of k-Step Latent (KSL)
PyTorch implementation of Soft Actor-Critic (SAC)
Proto-RL: Reinforcement Learning with Prototypical Representations
Control Ordinary Differential Equations with deep reinforcement learning
PyTorch Implementation of Visual GAIL in Atari Games
DrQ: Data regularized Q
This repository is a collection of widely used self-supervised auxiliary losses used for learning representations in reinforcement learning.
Base Mujoco Gymnasium environment for easily controlling any robot arm with operational space control. Built with dm-control PyMJCF for easy configuration.
A Multi-Task Dataset for Simulated Humanoid Control
Mujoco xml model for the Fetch Robotics Freight mobile base + Panda arm
A novel approach to solve Contextual Reinforcement Learning
Farama Gymnasium API Wrapper for the DeepMind Control Suite and DeepMind Robot Manipulation Tasks
Gymnasium integration for the DeepMind Control (DMC) suite
Add a description, image, and links to the dm-control topic page so that developers can more easily learn about it.
To associate your repository with the dm-control topic, visit your repo's landing page and select "manage topics."