Stable-baselines3 in c# with Torchsharp: Gym-based Reinforcement Learning framework

[GeorgeS2019]

Implementation of stable-baselines3 in c# with Torchsharp

Someone attempted in the past to port sb3 to c# using Torchsharp. Perhaps this rust-based approach could serve as a reference of achieving the goal.

Stable-baselines3 port written in rust using the burn library.

will-maclean/sb3-burn#36

[asieradzk]

Hi George. Have you been paying attention to RLMatrix? Tis about to reach feature parity of Rlib soon. You get shiny PPO and DQN-Rainbow out of the box with distributed reinforcement learning and as much optimization as I was able to squeeze out of C#. For networked stuff I am using SignalR with MessagePack serialisaton.

You're welcome to join the discussion, I am still debating what to use to serialise replay buffers...

I am not finished with networking stuff but I did a quick commit so you can have a look: