def test_tf_sparse(idx_shape):
import tensorflow as tf
from tensorflow import SparseTensor
from jina.types.ndarray.sparse.tensorflow import SparseNdArray
a = SparseTensor(indices=idx_shape[0], values=[1, 2, 3], dense_shape=idx_shape[1])
b = SparseNdArray()
b.value = a
np.testing.assert_equal(tf.sparse.to_dense(b.value).numpy(), tf.sparse.to_dense(a).numpy())
def test_dummy_numpy_sparse(shape):
a = np.random.random(shape)
a[a > 0.5] = 1
from jina.types.ndarray.sparse.numpy import SparseNdArray
b = SparseNdArray()
b.value = a
np.testing.assert_almost_equal(a, b.value)
def test_torch_sparse(idx_shape, transpose=False):
from jina.types.ndarray.sparse.pytorch import SparseNdArray
import torch
i = torch.LongTensor(idx_shape[0])
v = torch.FloatTensor([3, 4, 5])
a = torch.sparse.FloatTensor(i, v, torch.Size(idx_shape[1]))
b = SparseNdArray(transpose_indices=transpose)
b.value = a
np.testing.assert_equal(b.value.to_dense().numpy(), a.to_dense().numpy())
def test_scipy_sparse(sp_format):
from scipy.sparse import coo_matrix
from jina.types.ndarray.sparse.scipy import SparseNdArray
row = np.array([0, 3, 1, 0])
col = np.array([0, 3, 1, 2])
data = np.array([4, 5, 7, 9])
a = coo_matrix((data, (row, col)), shape=(4, 4))
dense_a = a.toarray()
b = SparseNdArray(sp_format=sp_format)
# write to proto
b.value = a
# read from proto
dense_b = b.value.toarray()
np.testing.assert_equal(dense_b, dense_a)