Is DeepMimic be trained using A3C or A2C?

[Zju-George]

A3C: aka Asynchronous Advantage Actor Critic

It uses MPI, so I wonder if DeepMimic be trained using A3C?

[xbpeng]

Neither, we are using PPO for training. The implementation with MPI is using synchronous updates, so it's more akin to A2C.