LemonPi opened this issue 5 years ago · 0 comments
Assuming dynamics, running cost, and terminal cost can operate on batches (K x <element size>), parallelize sampling and scoring trajectories
K x <element size>