For convenience, acceleration, and consistency, it is favourable to have all tensor operations written in PyTorch.
It is somewhat limited by PyTorch's ability to handle broadcasted linear algebra compared to NumPy, but
release 1.9 promises a better linalg module.
For convenience, acceleration, and consistency, it is favourable to have all tensor operations written in PyTorch.
It is somewhat limited by PyTorch's ability to handle broadcasted linear algebra compared to NumPy, but
release 1.9 promises a better linalg module.