rickstaa/stable-learning-control

Speed up torch algorithms using JIT

Opened this issue · 0 comments

The Pytorch versions of the algorithms can be speed up by using JIT. While doing this, you have to ensure we ignore the sample method since it is not yet JIT compatible (see pytorch/pytorch#29843 (comment)) and that we work around the incompatibility of torchscript with python type class attributes (see pytorch/pytorch#53753).