Any interest in this? There is, of course, a lot of need for scalable compute environments in Python, for machine learning. Using the (C) X10 runtime w bPython bindings could be very attractive. (Though we would want to consider extending runtime to work w/ K8…)
e.g. Ray supports distributed reinforcement learning libraries, and can be used to set up multi-agent simulations, e.g. for markets. But instead of simulating 20 agents, it would be good to simulate 20,000!! Ray: https://docs.ray.io/en/master/serve/key-concepts.html https://docs.google.com/document/d/1lAy0Owi-vPz2jEqBSaHNQcy2IBSDEHyXNOQZlGuj93c/preview Reinforcement Learning in Ray https://docs.ray.io/en/master/rllib.html Use of Ray for multi-agent market simulation https://arxiv.org/abs/1911.05892
_______________________________________________ X10-users mailing list X10-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/x10-users
