def sample(self, input, steps):
"""
Sample outputs from LM.
"""
inputs = [[onehot(self.input_dim, x) for x in input]]
for _ in range(steps):
target = self.compute(inputs)[0, -1].argmax()
input.append(target)
inputs[0].append(onehot(self.input_dim, target))
return input
def sample(self, input, steps):
"""
Sample outputs from LM.
"""
inputs = [[onehot(self.input_dim, x) for x in input]]
for _ in range(steps):
target = self.compute(inputs)[0, -1].argmax()
input.append(target)
inputs[0].append(onehot(self.input_dim, target))
return input
def prepare(self):
if not self.cached:
return
onehot_matrix = []
for i in xrange(self.vocab_size):
onehot_matrix.append(onehot(self.vocab_size, i))
onehot_matrix = np.array(onehot_matrix, dtype=FLOATX)
self.onehot_list = self.create_matrix(self.vocab_size, self.vocab_size, "onehot_list")
self.onehot_list.set_value(onehot_matrix)
def prepare(self):
if not self.cached:
return
onehot_matrix = []
for i in xrange(self.vocab_size):
onehot_matrix.append(onehot(self.vocab_size, i))
onehot_matrix = np.array(onehot_matrix, dtype=FLOATX)
self.onehot_list = self.create_matrix(self.vocab_size, self.vocab_size,
"onehot_list")
self.onehot_list.set_value(onehot_matrix)
def random_vector():
return onehot(VECTOR_SIZE, random.randint(0, VECTOR_SIZE - 1))
def random_vector():
return onehot(VECTOR_SIZE, random.randint(0, VECTOR_SIZE - 1))