rail-berkeley/softlearning
A deep reinforcement learning toolbox for training maximum entropy policies in continuous domains.

Velocity · 7d
+0.5
★ / day
Trend
→steady
star history
Softlearning is a reinforcement learning framework that implements the Soft Actor-Critic algorithm for training maximum entropy policies. It uses TensorFlow’s tf.keras modules for model classes like policies and value functions, and leverages Ray Tune for orchestrating and distributing experiments across cloud services. The framework targets continuous control tasks and integrates with MuJoCo for physics simulation.