Multiple Formats with Finch¶

Import¶

Let's set the backend and import sparse.

In [1]:

import os

os.environ["SPARSE_BACKEND"] = "Finch"

import sparse

import numpy as np

import os os.environ["SPARSE_BACKEND"] = "Finch" import sparse import numpy as np

/home/docs/checkouts/readthedocs.org/user_builds/sparse-nd/checkouts/latest/sparse/__init__.py:24: SparseFutureWarning: Changing back-ends is a development feature, please do not rely on it in production.
  warnings.warn(

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[1], line 5
      1 import os
      2 
      3 os.environ["SPARSE_BACKEND"] = "Finch"
      4 
----> 5 import sparse
      6 
      7 import numpy as np

File ~/checkouts/readthedocs.org/user_builds/sparse-nd/checkouts/latest/sparse/__init__.py:42
     39 del _backend_name
     41 if _BackendType.Finch == _BACKEND:
---> 42     from sparse.finch_backend import *  # noqa: F403
     43     from sparse.finch_backend import __all__
     44 elif _BackendType.MLIR == _BACKEND:

File ~/checkouts/readthedocs.org/user_builds/sparse-nd/checkouts/latest/sparse/finch_backend/__init__.py:6
      3 except ModuleNotFoundError as e:
      4     raise ImportError("Finch not installed. Run `pip install sparse[finch]` to enable Finch backend") from e
----> 6 from finch import *  # noqa: F403
      7 from finch import __all__ as __all__

AttributeError: module 'finch' has no attribute 'Reflector'

Perform Operations¶

Let's create two arrays.

In [2]:

rng = np.random.default_rng(42)  # Seed for reproducibility
a = sparse.random((3, 3), density=1 / 6, random_state=rng)
b = sparse.random((3, 3), density=1 / 6, random_state=rng)

rng = np.random.default_rng(42) # Seed for reproducibility a = sparse.random((3, 3), density=1 / 6, random_state=rng) b = sparse.random((3, 3), density=1 / 6, random_state=rng)

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[2], line 1
----> 1 rng = np.random.default_rng(42)  # Seed for reproducibility
      2 a = sparse.random((3, 3), density=1 / 6, random_state=rng)
      3 b = sparse.random((3, 3), density=1 / 6, random_state=rng)

NameError: name 'np' is not defined

Now let's matrix multiply them.

In [3]:

c = a @ b

c = a @ b

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[3], line 1
----> 1 c = a @ b

NameError: name 'a' is not defined

And view the result as a (dense) NumPy array.

In [4]:

c_dense = c.todense()

c_dense = c.todense()

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[4], line 1
----> 1 c_dense = c.todense()

NameError: name 'c' is not defined

Now let's do the same for other formats, and compare the results.

In [5]:

for format in ["coo", "csr", "csc", "dense"]:
    af = sparse.asarray(a, format=format)
    bf = sparse.asarray(b, format=format)
    cf = af @ bf
    np.testing.assert_array_equal(c_dense, cf.todense())

for format in ["coo", "csr", "csc", "dense"]: af = sparse.asarray(a, format=format) bf = sparse.asarray(b, format=format) cf = af @ bf np.testing.assert_array_equal(c_dense, cf.todense())

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
Cell In[5], line 2
      1 for format in ["coo", "csr", "csc", "dense"]:
----> 2     af = sparse.asarray(a, format=format)
      3     bf = sparse.asarray(b, format=format)
      4     cf = af @ bf
      5     np.testing.assert_array_equal(c_dense, cf.todense())

NameError: name 'sparse' is not defined