When training with a performer I get the following warning:
torch.qr is deprecated in favor of torch.linalg.qr and will be removed in a future PyTorch release.
The boolean parameter 'some' has been replaced with a string parameter 'mode'.
Q, R = torch.qr(A, some)
should be replaced with
Q, R = torch.linalg.qr(A, 'reduced' if some else 'complete') (Triggered internally at /pytorch/aten/src/ATen/native/BatchLinearAlgebra.cpp:1940.)
This patch fixes it by conditionally switching to the new invocation, while maintaining backwards compatibility.
When training with a performer I get the following warning:
This patch fixes it by conditionally switching to the new invocation, while maintaining backwards compatibility.