Hello,
I'm reproducing the reconstruction example in https://odlgroup.github.io/odl/guide/pdhg_guide.html, and I found out that the BroadcastOperator raises an Error related to the format and sparsity of the data when computing the L operator. I'm on python3.10.1, scipy1.10 and numpy1.23.5 on an OSX. I've tried downgrading scipy to the v1.8.1, which solves the issue with the Broadcasting operator, but other problems with the data types to construct the phantom appear instead. I've tried hacking scipy commenting the lines 127-129 in "/scipy/sparse/_sputils.py":
newdtype = np.dtype(dtype)
# Commented the section below to fix problem with ODL's broadcasting operator
#if newdtype == np.object_:
# raise ValueError(
# "object dtype is not supported by sparse matrices"
#)
And now the Broadcasting operator works.
Notice that scipy1.10.0 is installed with ODL.
Hello,
I'm reproducing the reconstruction example in https://odlgroup.github.io/odl/guide/pdhg_guide.html, and I found out that the BroadcastOperator raises an Error related to the format and sparsity of the data when computing the L operator. I'm on python3.10.1, scipy1.10 and numpy1.23.5 on an OSX. I've tried downgrading scipy to the v1.8.1, which solves the issue with the Broadcasting operator, but other problems with the data types to construct the phantom appear instead. I've tried hacking scipy commenting the lines 127-129 in "/scipy/sparse/_sputils.py":
And now the Broadcasting operator works.
Notice that scipy1.10.0 is installed with ODL.