search

SciSharp / TensorFlow.NET

.NET Standard bindings for Google's TensorFlow for developing, training and deploying Machine Learning models in C# and F#.
https://scisharp.github.io/tensorflow-net-docs
Apache License 2.0
3.26k stars 524 forks source link

Integrate with ml-agents #438

Closed Deep-Blue-2013 closed 4 years ago

Deep-Blue-2013 commented 5 years ago

If tf.net can be connected to this, it should be a lot easier. py often encounters some incompatibility problems, it is not easy to debug.

Unity Machine Learning Agents Toolkit https://github.com/Unity-Technologies/ml-agents

jorezy commented 5 years ago

Well, if Tensorflow.NET can support two deep reinforcement learning algorithms like Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC), I think Unity 3D developers will not have to rely on ml-agent, they can create more flexible AI agents in .NET with the support of TF.NET.

GeorgeS2019 commented 5 years ago

FYI Godot - the most active community 3D Engine for C# and UWP and AR needs support to integrate TensorFlow.NET

Godot also receive grant from Microsoft

Oceania2018 commented 5 years ago

Take a look at TF-Agents. A library for Reinforcement Learning in TensorFlow.

@Nucs What's the difference between OpenAI/Gym and TF-Agents ?

Nucs commented 5 years ago

OpenAI/Gym provides a set of standard environments to run RL algorithms against. TF-Agents uses OpenAI/Gym and other environment providers but it also provides its own algorithms and architecture for RL.

Agent == RL Algorithm

Oceania2018 commented 4 years ago

@GeorgeS2019 There is a PPO model implemented in TF.NET https://github.com/Oceania2018/unity-ml-agents.net.