`einsum` doesn't handle `dtype` and `optimize` kwargs
dcherian opened this issue · comments
Describe the bug
I tried to use np.einsum
on a combination of dask and sparse arrays. I get
TypeError: einsum() got an unexpected keyword argument 'dtype'
To Reproduce
import dask
import dask.array as da
data = dask.array.ones((5,5))
wgts = sparse.COO(coords=(0,1,2,3,4), data=[1, 1, 0.5, 1, 1], shape=(5,))
np.einsum('ij,i-> j', data, wgts)
Expected behavior
The output should match the pure numpy version
np.einsum('ij,i-> j', data.compute(), wgts.todense())
array([4.5, 4.5, 4.5, 4.5, 4.5])
System
- OS and version: macOS 13.3
sparse
version 0.14.0- NumPy version (
np.__version__
): 1.23.5 - Numba version (
numba.__version__
): 0.56.4
Additional context
Add any other context about the problem here.
@mtsokol Again, would this be something you'd be interested in?
Sure! I can take care of it.