Currently the code is using things like string formatting, rather than f-strings. Also, several of the tensors are being constantly detached and put on the CPU when it should be fine to allow them to be put on the GPU during computation. Change that.
Currently the code is using things like string formatting, rather than f-strings. Also, several of the tensors are being constantly detached and put on the CPU when it should be fine to allow them to be put on the GPU during computation. Change that.