lenskit.torch#

PyTorch utility functions.

Functions#

inference_mode(func)

Function decorator that puts PyTorch in inference mode.

sparse_row(mat, row)

Get a row of a sparse (CSR) tensor. This is needed because indexing a

safe_tensor(array)

Safely convert an array into a NumPy tensor. This includes copying it to

Module Contents#

lenskit.torch.inference_mode(func)#

Function decorator that puts PyTorch in inference mode.

lenskit.torch.sparse_row(mat, row)#

Get a row of a sparse (CSR) tensor. This is needed because indexing a tensor does not work in inference mode.

Parameters:
Return type:

torch.Tensor

lenskit.torch.safe_tensor(array)#

Safely convert an array into a NumPy tensor. This includes copying it to writable memory if necessary.

Return type:

torch.Tensor