def test_document_sparse_embedding(
scipy_sparse_matrix,
return_sparse_ndarray_cls_type,
return_scipy_class_type,
return_expected_type,
):
d = Document()
d.embedding = scipy_sparse_matrix
cls_type = None
sparse_kwargs = {}
if return_sparse_ndarray_cls_type == 'scipy':
from jina.types.ndarray.sparse.scipy import SparseNdArray
cls_type = SparseNdArray
sparse_kwargs['sp_format'] = return_scipy_class_type
elif return_sparse_ndarray_cls_type == 'torch':
from jina.types.ndarray.sparse.pytorch import SparseNdArray
cls_type = SparseNdArray
elif return_sparse_ndarray_cls_type == 'tf':
from jina.types.ndarray.sparse.tensorflow import SparseNdArray
cls_type = SparseNdArray
embedding = d.get_sparse_embedding(sparse_ndarray_cls_type=cls_type,
**sparse_kwargs)
assert embedding is not None
assert isinstance(embedding, return_expected_type)
if return_sparse_ndarray_cls_type == 'torch':
assert embedding.is_sparse
def test_document_sparse_embedding(
scipy_sparse_matrix,
return_sparse_ndarray_cls_type,
return_scipy_class_type,
return_expected_type,
field,
):
d = Document()
setattr(d, field, scipy_sparse_matrix)
cls_type = None
sparse_kwargs = {}
if return_sparse_ndarray_cls_type == 'scipy':
from jina.types.ndarray.sparse.scipy import SparseNdArray
cls_type = SparseNdArray
sparse_kwargs['sp_format'] = return_scipy_class_type
elif return_sparse_ndarray_cls_type == 'torch':
from jina.types.ndarray.sparse.pytorch import SparseNdArray
cls_type = SparseNdArray
elif return_sparse_ndarray_cls_type == 'tf':
from jina.types.ndarray.sparse.tensorflow import SparseNdArray
cls_type = SparseNdArray
if field == 'blob':
field_sparse = d.get_sparse_blob(sparse_ndarray_cls_type=cls_type,
**sparse_kwargs)
elif field == 'embedding':
field_sparse = d.get_sparse_embedding(sparse_ndarray_cls_type=cls_type,
**sparse_kwargs)
assert field_sparse is not None
assert isinstance(field_sparse, return_expected_type)
if return_sparse_ndarray_cls_type == 'torch':
assert field_sparse.is_sparse
if return_sparse_ndarray_cls_type == 'scipy':
np.testing.assert_equal(field_sparse.todense(),
scipy_sparse_matrix.todense())
elif return_sparse_ndarray_cls_type == 'torch':
np.testing.assert_equal(field_sparse.to_dense().numpy(),
scipy_sparse_matrix.todense())
elif return_scipy_class_type == 'tf':
np.testing.assert_equal(
tf.sparse.to_dense(field_sparse).numpy(),
scipy_sparse_matrix.todense())
