What is the current roadmap?

SciSharp / Gym.NET

openai/gym's popular toolkit for developing and comparing reinforcement learning algorithms port to C#.

Apache License 2.0

119 stars 20 forks source link

What is the current roadmap? #1

Open Kavignon opened 4 years ago

Kavignon commented 4 years ago

What’s the current roadmap of the project? Is it possible to contribute?

Nucs commented 4 years ago

Hey @Kavignon, we accept contributions. A "roadmap" is available on the main readme.md (as TODO). In a nutshell, the library provides most of the behind-the-scenes architecture and currently a single environment cartpole-v1 written in pure C#. Having the most of the architecture done it is easy to "wrap" custom environments using openai/gym's pattern.

Currently we are missing:

Environments - detailed in the readme.md.
Examples with implementations of algorithms like DQN, PPO, DDQN using our other libraries TF.NET and NumSharp

dariooo512 commented 4 years ago

Hi @Nucs, i'm working on a sample of a DQN using NeuralNetwork.NET.

Could it be a valid contribution to the project? (asking this because you only mentioned TF.NET and NumSharp)

Nucs commented 4 years ago

Hey @dariooo512, NeuralNetwork.NET (which is also part of SciSharp's ecosystem as can be seen in our website scisharpstack.org) is definitely a valid candidate for an example. I'll be happy to add it as an example in a separate project (examples/Gym.Examples) inside this repository.

Gimly commented 3 years ago

@Nucs What is the current state of this project? The last commit has been pushed nearly a year ago and there is an unmerged PR that is open without any comments for quite some times.

I'd be interested in contributing, but I'd like to be sure that the project is still alive.

Nucs commented 3 years ago

Unmaintained due to lacking API from TF.NET, you can speak with @Oceania2018 if you think you have what it takes to take over.