It appears that partial_unfold works using sparse tensors, but it is not clear in the documentation

[mdarmstr]

Creating a sparse tensor using the sparse backend. Works as expected with unfold, and with partial_unfold.

import numpy as np
import sparse
from tensorly.contrib.sparse import tensor, unfold
from tensorly.base import partial_unfold

X = sparse.COO(np.arange(1152).reshape((24,3,2,8)))
X = sparse.COO(X.coords, X.data, shape=(100, 100, 100, 100))
X = tensor(X, dtype='float')
print(unfold(X, 2)) # mode-1 unfolding, from tensorly.contrib.sparse
print(partial_unfold(X,0,1,0))

output:
<COO: shape=(100, 1000000), dtype=float64, nnz=1151, fill_value=0.0>
<COO: shape=(100, 100, 10000), dtype=float64, nnz=1151, fill_value=0.0>

Linux-5.19.0-46-generic-x86_64-with-glibc2.35Python 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0]
NumPy 1.24.4
SciPy 1.11.1
TensorLy 0.8.1