def test_save_and_load(self):
"""Try to save and load encoder."""
vocabulary = Vocabulary()
vocabulary.add_word("a")
vocabulary.add_word("b")
checkpoint_file = tempfile.NamedTemporaryFile(delete=False)
checkpoint_file.close()
encoder = SentenceEncoder(
"enc", Vocabulary(), "data_id", 10, 20, 30,
save_checkpoint=checkpoint_file.name,
load_checkpoint=checkpoint_file.name)
encoders_variables = tf.get_collection(
tf.GraphKeys.VARIABLES, scope="enc")
sess_1 = tf.Session()
sess_1.run(tf.initialize_all_variables())
encoder.save(sess_1)
sess_2 = tf.Session()
sess_2.run(tf.initialize_all_variables())
encoder.load(sess_2)
values_in_sess_1 = sess_1.run(encoders_variables)
values_in_sess_2 = sess_2.run(encoders_variables)
self.assertTrue(
all(np.all(v1 == v2) for v1, v2 in
zip(values_in_sess_1, values_in_sess_2)))
os.remove(checkpoint_file.name)
class Word2Vec:
def __init__(self, path: str, encoding: str = "utf-8") -> None:
"""Load the word2vec file."""
check_argument_types()
# Create the vocabulary object, load the words and vectors from the
# file
self.vocab = Vocabulary()
embedding_vectors = [] # type: List[np.ndarray]
with open(path, encoding=encoding) as f_data:
header = next(f_data)
emb_size = int(header.split()[1])
# Add zero embeddings for padding, start, and end token
embedding_vectors.append(np.zeros(emb_size))
embedding_vectors.append(np.zeros(emb_size))
embedding_vectors.append(np.zeros(emb_size))
# Add placeholder for embedding of the unknown symbol
embedding_vectors.append(None)
for line in f_data:
fields = line.split()
word = fields[0]
vector = np.fromiter((float(x) for x in fields[1:]),
dtype=np.float)
assert vector.shape[0] == emb_size
# Embedding of unknown token should be at index 3 to match the
# vocabulary implementation
if is_special_token(word):
embedding_vectors[SPECIAL_TOKENS.index(word)] = vector
else:
self.vocab.add_word(word)
embedding_vectors.append(vector)
assert embedding_vectors[3] is not None
assert emb_size is not None
self.embedding_matrix = np.stack(embedding_vectors)
@property
def vocabulary(self) -> Vocabulary:
"""Get a vocabulary object generated from this word2vec instance."""
return self.vocab
@property
def embeddings(self) -> np.ndarray:
"""Get the embedding matrix."""
return self.embedding_matrix
def test_save_and_load(self):
"""Try to save and load encoder."""
vocabulary = Vocabulary()
vocabulary.add_word("a")
vocabulary.add_word("b")
checkpoint_file = tempfile.NamedTemporaryFile(delete=False)
checkpoint_file.close()
encoder = SentenceEncoder(name="enc",
vocabulary=Vocabulary(),
data_id="data_id",
embedding_size=10,
rnn_size=20,
max_input_len=30,
save_checkpoint=checkpoint_file.name,
load_checkpoint=checkpoint_file.name)
encoder.input_sequence.register_input()
# NOTE: This assert needs to be here otherwise the model has
# no parameters since the sentence encoder is initialized lazily
self.assertIsInstance(encoder.temporal_states, tf.Tensor)
encoders_variables = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES,
scope="enc")
sess_1 = tf.Session()
sess_1.run(tf.global_variables_initializer())
encoder.save(sess_1)
sess_2 = tf.Session()
sess_2.run(tf.global_variables_initializer())
encoder.load(sess_2)
values_in_sess_1 = sess_1.run(encoders_variables)
values_in_sess_2 = sess_2.run(encoders_variables)
self.assertTrue(
all(
np.all(v1 == v2)
for v1, v2 in zip(values_in_sess_1, values_in_sess_2)))
os.remove(checkpoint_file.name)
def test_reuse(self):
vocabulary = Vocabulary()
vocabulary.add_word("a")
vocabulary.add_word("b")
seq1 = EmbeddedSequence(name="seq1",
vocabulary=vocabulary,
data_id="id",
embedding_size=10)
seq1.register_input()
seq2 = EmbeddedSequence(name="seq2",
vocabulary=vocabulary,
embedding_size=10,
data_id="id")
seq2.register_input()
seq3 = EmbeddedSequence(name="seq3",
vocabulary=vocabulary,
data_id="id",
embedding_size=10,
reuse=seq1)
seq3.register_input()
# blessing
self.assertIsNotNone(seq1.embedding_matrix)
self.assertIsNotNone(seq2.embedding_matrix)
self.assertIsNotNone(seq3.embedding_matrix)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
params = sess.run((seq1.embedding_matrix, seq2.embedding_matrix,
seq3.embedding_matrix))
with self.assertRaises(AssertionError):
assert_array_equal(params[0], params[1])
assert_array_equal(params[0], params[2])
