I have a question. Does PyTorch support TensorRT (NVIDIA TensorRT)? Or does any future plan for support it?
I don’t know how well TensorRT supports PyTorch ATM, but I know that Nvidia has at least one person working on PyTorch / Torch / etc support. An anecdotal bit of weirdness that he likes mentioning is that PyTorch and Torch have incompatible RNN weight representations 0.o
Check this out (converting and running a pytorch model in a tensorRT engine) -->