def generate_text(model, seed, length=512, top_n=None):
"""
generates text of specified length from trained model
with given seed character sequence.
"""
logger.info("generating %s characters from top %s choices.", length, top_n)
logger.info('generating with seed: "%s".', seed)
generated = seed
encoded = encode_text(seed, get_CHAR2ID())
model.reset_states()
for idx in encoded[:-1]:
x = np.array([[idx]])
# input shape: (1, 1)
# set internal states
model.predict(x)
next_index = encoded[-1]
for i in range(length):
x = np.array([[next_index]])
# input shape: (1, 1)
probs = model.predict(x)
# output shape: (1, 1, vocab_size)
next_index = sample_from_probs(probs.squeeze(), top_n)
# append to sequence
generated += get_ID2CHAR()[next_index]
logger.info("generated text: \n%s\n", generated)
return generated
def generate_text(model, seed, length=512, top_n=10):
"""
generates text of specified length from trained model
with given seed character sequence.
"""
logger.info("generating %s characters from top %s choices.", length, top_n)
logger.info('generating with seed: "%s".', seed)
generated = seed
encoded = encode_text(seed).astype(np.int32)
model.predictor.reset_state()
with chainer.using_config("train", False), chainer.no_backprop_mode():
for idx in encoded[:-1]:
x = Variable(np.array([idx]))
# input shape: [1]
# set internal states
model.predictor(x)
next_index = encoded[-1]
for i in range(length):
x = Variable(np.array([next_index], dtype=np.int32))
# input shape: [1]
probs = F.softmax(model.predictor(x))
# output shape: [1, vocab_size]
next_index = sample_from_probs(probs.data.squeeze(), top_n)
# append to sequence
generated += ID2CHAR[next_index]
logger.info("generated text: \n%s\n", generated)
return generated
def generate_text(model, seed, length=512, top_n=2):
"""
Generates text of specified length from trained model with given seed (e.g. the prefix string).
"""
logger.info("generating %s characters from top %s choices.", length, top_n)
logger.info('generating with seed: "%s".', seed)
generated = seed
encoded = encode_text(seed)
model.reset_states()
for idx in encoded[:-1]:
x = np.array([[idx]])
# Input shape: (1, 1)
# Set internal states
model.predict(x)
next_index = encoded[-1]
for i in range(length):
x = np.array([[next_index]])
# Input shape: (1, 1)
probs = model.predict(x)
# Output shape: (1, 1, vocab_size)
next_index = sample_from_probs(probs.squeeze(), top_n)
# Append to sequence
if ID2CHAR[next_index] in [".", "!", "?"]:
generated += ID2CHAR[next_index]
break
elif ID2CHAR[next_index] == "\n":
break
generated += ID2CHAR[next_index]
logger.info("generated text: \n%s\n", generated)
return generated
def generate_text(model, sess, seed, length=512, top_n=10):
"""
generates text of specified length from trained model
with given seed character sequence.
"""
logger.info("generating %s characters from top %s choices.", length, top_n)
logger.info('generating with seed: "%s".', seed)
generated = seed
encoded = encode_text(seed)
x = np.expand_dims(encoded[:-1], 0)
# input shape: [1, seq_len]
# get rnn state due to seed sequence
state = sess.run(model["output_state"], feed_dict={model["X"]: x})
next_index = encoded[-1]
for i in range(length):
x = np.array([[next_index]])
# input shape: [1, 1]
feed_dict = {model["X"]: x, model["input_state"]: state}
probs, state = sess.run([model["probs"], model["output_state"]],
feed_dict=feed_dict)
# output shape: [1, 1, vocab_size]
next_index = sample_from_probs(probs.squeeze(), top_n)
# append to sequence
generated += ID2CHAR[next_index]
logger.info("generated text: \n%s\n", generated)
return generated
def generate_text(model, seed, length=512, top_n=10):
"""
generates text of specified length from trained model
with given seed character sequence.
"""
logger.info("generating %s characters from top %s choices.", length, top_n)
logger.info('generating with seed: "%s".', seed)
generated = seed
encoded = mx.nd.array(encode_text(seed))
seq_len = encoded.shape[0]
x = F.expand_dims(encoded[:seq_len - 1], 1)
# input shape: [seq_len, 1]
state = model.begin_state()
# get rnn state due to seed sequence
_, state = model(x, state)
next_index = encoded[seq_len - 1].asscalar()
for i in range(length):
x = mx.nd.array([[next_index]])
# input shape: [1, 1]
logit, state = model(x, state)
# output shape: [1, vocab_size]
probs = F.softmax(logit)
next_index = sample_from_probs(probs.asnumpy().squeeze(), top_n)
# append to sequence
generated += ID2CHAR[next_index]
logger.info("generated text: \n%s\n", generated)
return generated
