Training without Python

[NeuralNotworks]

Please let me know if I’m asking in the wrong channel.

I wonder if you have considered adding a feature of training ML models inside the gameplay? AFAIK, the current ML-Agent package allows us to use Unity for virtual environments, but the training is taken care of by Python. If training is possible with only, say C#, then we can develop games in which we train our own AI’s.

Or is this already possible with some other packages?

[macsimilian]

ML Agents uses TensorFlow under the hood. ML Agents entirely inside Unity could potentially be possible since there is a JavaScript version of TensorFlow, and Unity uses JavaScript.

[NeuralNotworks]

@macsimilian

ML Agents uses TensorFlow under the hood. ML Agents entirely inside Unity could potentially be possible since there is a JavaScript version of TensorFlow, and Unity uses JavaScript.

Thanks for the answer, macsimilian! This is indeed possible! A quick search showed that there were already people did this, such as the in-browser demo in this blog post.