PhilJd/contiguous_pytorch_params

Accelerate training by storing parameters in one contiguous chunk of memory.

Python