def generate_model():
text = read_text("../data/shakespeare")
vocab = read_corpus(text)
char_idx = {u: i for i, u in enumerate(vocab)}
idx2char = np.array(vocab)
dataset = start_model(char_idx, text)
# build model
model = build_model(vocab_size=len(vocab),
embedding_dim=300,
rnn_units=1024,
batch_size=128)
model.compile(optimizer='adam', loss=loss)
epochs = 1
checkpoint_prefix = os.path.join("./train", "ckpt_{epoch}")
# callback function called at the end of epoch training
checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
filepath=checkpoint_prefix, save_weights_only=True)
new_model = build_model(vocab_size=len(vocab),
embedding_dim=300,
rnn_units=1024,
batch_size=1)
return new_model, char_idx, idx2char
def test_start_model(self):
text = read_text("../data/shakespeare/")
vocab = read_corpus(text)
char_idx = {u: i for i, u in enumerate(vocab)}
test = start_model(char_idx, text)
self.assertNotEqual(test, None)
def setUpClass(cls):
cls.grail = read_text('data/grail.txt')
cls.emma = read_text('data/emma.txt')
cls.wsj = read_text('data/wsj')
def setUpClass(cls):
cls.text = read_text('data/grail.txt')
cls.vocab = Vocabulary(cls.text)
def test_read_corpus(self):
text = read_text("../data/shakespeare/")
vocab = read_corpus(text)
self.assertNotEqual(vocab, "")
def test_read_text(self):
text = read_text("../data/shakespeare/")
self.assertNotEqual(text, "")
self.assertRegex(text, "[A-Za-z]*")