def vectors(request, dataset):
"""
Generate random word vectors
Params: dimensionality
"""
output_vec = normalize_rows(np.random.randn(dataset.dim, request.param))
input_vec = normalize_rows(np.random.randn(1, request.param))
return input_vec, output_vec
def test_normalize_rows():
first = normalize_rows(np.array([[3.0, 4.0], [1.0, 2.0]]))
second = np.array([[0.6, 0.8], [0.4472, 0.8944]])
assert_close(
first,
second
)
def dummy_vectors(dataset):
dim = 10
weights = 0.1 * np.random.randn(dim, 5)
features = np.zeros((dataset.dim, dim))
labels = np.zeros((dataset.dim,), dtype=np.int32)
word_vectors = normalize_rows(np.random.randn(dataset.dim, dim))
for i in xrange(dataset.dim):
words, labels[i] = dataset.get_random_train_sentence()
features[i, :] = get_sentence_feature(dataset.tokens, word_vectors, words)
return weights, features, labels
