rail-berkeley / rlkit

Collection of reinforcement learning algorithms
MIT License
2.45k stars 550 forks source link

Introducing possibility to change the standard alpha parameter for SAC #110

Closed ksluck closed 4 years ago

ksluck commented 4 years ago

It would be nice if we could set our own alpha parameter if we choose not to use the automatic parameter tuning. At the moment, the SAC class sets alpha always to one.

The proposed change introduces a new parameter for SAC, alpha, with the default value of 1.0 which should not lead to any issues or conflicts with prior versions.