PyTorch extension library bringing the isin() function similar to numpy.isin() to PyTorch tensors, supporting both CPU and CUDA.
Currently, you can only install the extension by compiling it from the source.
Make sure that torch (tested with version 1.7.0) is installed. If you want CUDA support, make sure you got the CUDA development kit setup with the same version you installed torch with. Make sure that your $PATH, and CUDA_PATH or CPATH are setup correctly.
$ python -c "import torch; print(torch.__version__)"
> 1.7.0
$ echo $PATH
> /usr/local/cuda/bin:...
$ echo $CUDA_PATH
> /usr/local/cuda/include:...Then run:
$ python setup.py install
> ...isin(elements: torch.Tensor, test_elements: torch.Tensor, invert: bool = False)
Returns a boolean array of the same shape as elements that is True where an element of elements is in test_elements and False otherwise.
Note: This function is equivalent to computing the element-wise Python keyword isin(a, b) for each element in the input.
import torch
from torch_isin import isin
x = torch.tensor([0, 1, 1, 2])
y = torch.tensor([0, 2])
out = isin(x, y)print(out)
tensor([True, False, False, True])$ python setup.py test
> ...