search

Unity-Technologies / ml-agents

The Unity Machine Learning Agents Toolkit (ML-Agents) is an open-source project that enables games and simulations to serve as environments for training intelligent agents using deep reinforcement learning and imitation learning.
https://unity.com/products/machine-learning-agents
Other
17.12k stars 4.15k forks source link

Externally feeding SAC experience buffer #5722

Closed AleksanderWit closed 2 years ago

AleksanderWit commented 2 years ago

Is your feature request related to a problem? Please describe. I am looking into making a multiplayer game or an otherwise online game that would include MLAgent models that would train either continously or periodically with breaks. The big issue would be to provide all those player interactions to the model being trained in some effective form or another.

Describe the solution you'd like These are the solutions I can think of right now with my limited knowledge:

Describe alternatives you've considered Using separate agents to directly play in the various player worlds. (I believe this is the only one I have a very vague idea for how to acheive right now. I could potentially feed information from those online sessions ran on other computers directly to a Unity environment the AI is traing on. This sounds very network intensive and slow though.)

Additional context I am personally very new to the practical and execution world of machine learning in general. With Unity ML Agents even more so. There may be obvious things I may be missing that either make all of this infeasible or all of this already easily doable.

miguelalonsojr commented 2 years ago

Thanks for your feature request! We've logged this feature request internally as MLA-2488.

github-actions[bot] commented 2 years ago

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.