def initial_setup(data_corpus):
metadata, idx_q, idx_a = data.load_data(PATH='data/{}/'.format(data_corpus))
(trainX, trainY), (testX, testY), (validX, validY) = data.split_dataset(idx_q, idx_a)
trainX = tl.prepro.remove_pad_sequences(trainX.tolist())
trainY = tl.prepro.remove_pad_sequences(trainY.tolist())
testX = tl.prepro.remove_pad_sequences(testX.tolist())
testY = tl.prepro.remove_pad_sequences(testY.tolist())
validX = tl.prepro.remove_pad_sequences(validX.tolist())
validY = tl.prepro.remove_pad_sequences(validY.tolist())
return metadata, trainX, trainY, testX, testY, validX, validY
def __init__(self):
data_corpus = "twitter"
metadata, idx_q, idx_a = data.load_data(
PATH='data/{}/'.format(data_corpus))
src_vocab_size = len(metadata['idx2w']) # 8002 (0~8001)
emb_dim = 1024
word2idx = metadata['w2idx'] # dict word 2 index
idx2word = metadata['idx2w'] # list index 2 word
unk_id = word2idx['unk'] # 1
pad_id = word2idx['_'] # 0
start_id = src_vocab_size # 8002
end_id = src_vocab_size + 1 # 8003
word2idx.update({'start_id': start_id})
word2idx.update({'end_id': end_id})
idx2word = idx2word + ['start_id', 'end_id']
src_vocab_size = tgt_vocab_size = src_vocab_size + 2
# num_epochs = 5
vocabulary_size = src_vocab_size
decoder_seq_length = 20
self.unk_id = unk_id
self.pad_id = pad_id
self.start_id = start_id
self.end_id = end_id
self.word2idx = word2idx
self.idx2word = idx2word
self.model_ = Seq2seq(
decoder_seq_length=decoder_seq_length,
cell_enc=tf.keras.layers.GRUCell,
cell_dec=tf.keras.layers.GRUCell,
n_layer=3,
n_units=256,
embedding_layer=tl.layers.Embedding(
vocabulary_size=vocabulary_size, embedding_size=emb_dim),
)
load_weights = tl.files.load_npz(name='model.npz')
tl.files.assign_weights(load_weights, self.model_)
def load_data(path):
metadata, idx_q, idx_a = data.load_data(path)
(trainX, trainY), (testX,
testY), (validX,
validY) = data.split_dataset(idx_q, idx_a)
trainX = trainX.tolist()
trainY = trainY.tolist()
testX = testX.tolist()
testY = testY.tolist()
validX = validX.tolist()
validY = validY.tolist()
trainX = tl.prepro.remove_pad_sequences(trainX)
trainY = tl.prepro.remove_pad_sequences(trainY)
testX = tl.prepro.remove_pad_sequences(testX)
testY = tl.prepro.remove_pad_sequences(testY)
validX = tl.prepro.remove_pad_sequences(validX)
validY = tl.prepro.remove_pad_sequences(validY)
return trainX, trainY, testX, testY, validX, validY, metadata
#! /usr/bin/python
# -*- coding: utf8 -*-
import tensorflow as tf
import tensorlayer as tl
from tensorlayer.layers import *
import tensorflow as tf
import numpy as np
import time
###============= prepare data
from data.twitter import data
metadata, idx_q, idx_a = data.load_data('data/twitter/') # Twitter
# from data.cornell_corpus import data
# metadata, idx_q, idx_a = data.load_data(PATH='data/cornell_corpus/') # Cornell Moive
(trainX, trainY), (testX, testY), (validX, validY) = data.split_dataset(idx_q, idx_a)
trainX = trainX.tolist()
trainY = trainY.tolist()
testX = testX.tolist()
testY = testY.tolist()
validX = validX.tolist()
validY = validY.tolist()
trainX = tl.prepro.remove_pad_sequences(trainX)
trainY = tl.prepro.remove_pad_sequences(trainY)
testX = tl.prepro.remove_pad_sequences(testX)
testY = tl.prepro.remove_pad_sequences(testY)
validX = tl.prepro.remove_pad_sequences(validX)
validY = tl.prepro.remove_pad_sequences(validY)
References ---------- http://suriyadeepan.github.io/2016-12-31-practical-seq2seq/ """ import tensorflow as tf import tensorlayer as tl from tensorlayer.layers import * import tensorflow as tf import numpy as np import time ###============= prepare data from data.twitter import data metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/') # Twitter # from data.cornell_corpus import data # metadata, idx_q, idx_a = data.load_data(PATH='data/cornell_corpus/') # Cornell Moive (trainX, trainY), (testX, testY), (validX, validY) = data.split_dataset(idx_q, idx_a) trainX = trainX.tolist() trainY = trainY.tolist() testX = testX.tolist() testY = testY.tolist() validX = validX.tolist() validY = validY.tolist() trainX = tl.prepro.remove_pad_sequences(trainX) trainY = tl.prepro.remove_pad_sequences(trainY) testX = tl.prepro.remove_pad_sequences(testX) testY = tl.prepro.remove_pad_sequences(testY)
def main():
metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/') # Twitter
# from data.cornell_corpus import data
# metadata, idx_q, idx_a = data.load_data(PATH='data/cornell_corpus/') # Cornell Moive
(trainX, trainY), (testX,
testY), (validX,
validY) = data.split_dataset(idx_q, idx_a)
trainX = trainX.tolist()
trainY = trainY.tolist()
testX = testX.tolist()
testY = testY.tolist()
validX = validX.tolist()
validY = validY.tolist()
trainX = tl.prepro.remove_pad_sequences(trainX)
trainY = tl.prepro.remove_pad_sequences(trainY)
testX = tl.prepro.remove_pad_sequences(testX)
testY = tl.prepro.remove_pad_sequences(testY)
validX = tl.prepro.remove_pad_sequences(validX)
validY = tl.prepro.remove_pad_sequences(validY)
###============= parameters
xseq_len = len(trainX) #.shape[-1]
yseq_len = len(trainY) #.shape[-1]
assert xseq_len == yseq_len
batch_size = 32
n_step = int(xseq_len / batch_size)
xvocab_size = len(metadata['idx2w']) # 8002 (0~8001)
emb_dim = 1024
global w2idx
global idx2w
global encode_seqs2
global decode_seqs2
global start_id
global end_id
w2idx = metadata['w2idx'] # dict word 2 index
idx2w = metadata['idx2w'] # list index 2 word
unk_id = w2idx['unk'] # 1
pad_id = w2idx['_'] # 0
start_id = xvocab_size # 8002
end_id = xvocab_size + 1 # 8003
w2idx.update({'start_id': start_id})
w2idx.update({'end_id': end_id})
idx2w = idx2w + ['start_id', 'end_id']
xvocab_size = yvocab_size = xvocab_size + 2
""" A data for Seq2Seq should look like this:
input_seqs : ['how', 'are', 'you', '<PAD_ID'>]
decode_seqs : ['<START_ID>', 'I', 'am', 'fine', '<PAD_ID'>]
target_seqs : ['I', 'am', 'fine', '<END_ID>', '<PAD_ID'>]
target_mask : [1, 1, 1, 1, 0]
"""
print("encode_seqs", [idx2w[id] for id in trainX[10]])
target_seqs = tl.prepro.sequences_add_end_id([trainY[10]],
end_id=end_id)[0]
# target_seqs = tl.prepro.remove_pad_sequences([target_seqs], pad_id=pad_id)[0]
print("target_seqs", [idx2w[id] for id in target_seqs])
decode_seqs = tl.prepro.sequences_add_start_id([trainY[10]],
start_id=start_id,
remove_last=False)[0]
# decode_seqs = tl.prepro.remove_pad_sequences([decode_seqs], pad_id=pad_id)[0]
print("decode_seqs", [idx2w[id] for id in decode_seqs])
target_mask = tl.prepro.sequences_get_mask([target_seqs])[0]
print("target_mask", target_mask)
print(len(target_seqs), len(decode_seqs), len(target_mask))
###============= model
global net_rnn
def model(encode_seqs, decode_seqs, is_train=True, reuse=False):
with tf.variable_scope("model", reuse=reuse):
# for chatbot, you can use the same embedding layer,
# for translation, you may want to use 2 seperated embedding layers
with tf.variable_scope("embedding") as vs:
net_encode = EmbeddingInputlayer(inputs=encode_seqs,
vocabulary_size=xvocab_size,
embedding_size=emb_dim,
name='seq_embedding')
vs.reuse_variables()
tl.layers.set_name_reuse(True)
net_decode = EmbeddingInputlayer(inputs=decode_seqs,
vocabulary_size=xvocab_size,
embedding_size=emb_dim,
name='seq_embedding')
net_rnn = Seq2Seq(
net_encode,
net_decode,
cell_fn=tf.contrib.rnn.BasicLSTMCell,
n_hidden=emb_dim,
initializer=tf.random_uniform_initializer(-0.1, 0.1),
encode_sequence_length=retrieve_seq_length_op2(encode_seqs),
decode_sequence_length=retrieve_seq_length_op2(decode_seqs),
initial_state_encode=None,
dropout=(0.5 if is_train else None),
n_layer=3,
return_seq_2d=True,
name='seq2seq')
net_out = DenseLayer(net_rnn,
n_units=xvocab_size,
act=tf.identity,
name='output')
return net_out, net_rnn
# model for training
encode_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="encode_seqs")
decode_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="decode_seqs")
target_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="target_seqs")
target_mask = tf.placeholder(
dtype=tf.int64, shape=[batch_size, None],
name="target_mask") # tl.prepro.sequences_get_mask()
net_out, _ = model(encode_seqs, decode_seqs, is_train=True, reuse=False)
# model for inferencing
encode_seqs2 = tf.placeholder(dtype=tf.int64,
shape=[1, None],
name="encode_seqs")
decode_seqs2 = tf.placeholder(dtype=tf.int64,
shape=[1, None],
name="decode_seqs")
net, net_rnn = model(encode_seqs2,
decode_seqs2,
is_train=False,
reuse=True)
global y
y = tf.nn.softmax(net.outputs)
# loss for training
# print(net_out.outputs) # (?, 8004)
# print(target_seqs) # (32, ?)
# loss_weights = tf.ones_like(target_seqs, dtype=tf.float32)
# loss = tf.contrib.legacy_seq2seq.sequence_loss(net_out.outputs, target_seqs, loss_weights, yvocab_size)
loss = tl.cost.cross_entropy_seq_with_mask(logits=net_out.outputs,
target_seqs=target_seqs,
input_mask=target_mask,
return_details=False,
name='cost')
net_out.print_params(False)
lr = 0.0001
train_op = tf.train.AdamOptimizer(learning_rate=lr).minimize(loss)
# Truncated Backpropagation for training (option)
# max_grad_norm = 30
# grads, _ = tf.clip_by_global_norm(tf.gradients(loss, net_out.all_params),max_grad_norm)
# optimizer = tf.train.GradientDescentOptimizer(lr)
# train_op = optimizer.apply_gradients(zip(grads, net_out.all_params))
# sess = tf.InteractiveSession()
global sess
sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False))
tl.layers.initialize_global_variables(sess)
tl.files.load_and_assign_npz(sess=sess, name='n.npz', network=net)
run(host='0.0.0.0', port=40026)
class AI:
metadata, idx_q, idx_a = data.load_data(PATH='data/{}/'.format("twitter"))
src_vocab_size = len(metadata['idx2w'])
emb_dim = 1024
word2idx = metadata['w2idx']
idx2word = metadata['idx2w']
unk_id = word2idx['unk']
pad_id = word2idx['_']
start_id = src_vocab_size
end_id = src_vocab_size + 1
word2idx.update({'start_id': start_id})
word2idx.update({'end_id': end_id})
idx2word = idx2word + ['start_id', 'end_id']
src_vocab_size = src_vocab_size + 2
vocabulary_size = src_vocab_size
decoder_seq_length = 20
# Creates an instance of the AI with it's name being passed from a name-generating function
def __init__(self):
# preProcessor is only utilised here to avoid duplicating string2array
self.preProcessor = DataPreProcessor("")
self.model_ = Seq2seq(
decoder_seq_length=self.decoder_seq_length,
cell_enc=tf.keras.layers.GRUCell,
cell_dec=tf.keras.layers.GRUCell,
n_layer=3,
n_units=256,
embedding_layer=tl.layers.Embedding(
vocabulary_size=self.vocabulary_size,
embedding_size=self.emb_dim),
)
load_weights = tl.files.load_npz(name="data/model.npz")
tl.files.assign_weights(load_weights, self.model_)
# Take a name, create a mood
def initialise(self, name):
self.name = name
feelings = [
"good", "well", "great", "grand", "excellent", "ecstatic", "happy",
"sad", "annoyed", "frustrated", "angry", "tired", "okay", "alright"
]
self.feel = feelings[randint(0, len(feelings) - 1)]
# Handle the creation of a response from the given input
def respond(self, seed, number):
simpleStart = self.simpleResponse(seed)
self.model_.eval()
seed_id = [self.word2idx.get(w, self.unk_id) for w in seed.split(" ")]
sentence_id = self.model_(inputs=[[seed_id]],
seq_length=20,
start_token=self.start_id,
top_n=number)
sentence = []
for w_id in sentence_id[0]:
w = self.idx2word[w_id]
if w == 'end_id':
break
sentence = sentence + [w]
# A catch all just in case there are no responses, but we have yet to find an input to trigger this
if sentence == []:
sentence = [
"I'm", "sorry,", "I", "just", "don't", "quite", "understand",
"what", "you're", "asking..."
]
return simpleStart + sentence
# Handle simple questions that the AI is less than optimal at answering
def simpleResponse(self, input):
sentence = []
input = self.preProcessor.string2Array(input)
tally = [0, 0, 0]
greetings = [
"hello", "hi", "greetings", "salutations", "hey", "yo", "howdy"
]
names = [["what", "who"], ["is", "are"], ["you", "your"], ["name"]]
wellbeing = [["how"], ["do", "are"], ["you"],
["doing", "feeling", "feel"]]
# Tallying key words in the user query to determine if certain questions were being asked
for x in input:
for y in range(len(greetings)):
if x == greetings[y]:
tally[0] = 1
break
for y in range(len(names)):
for z in names[y]:
if x is z:
tally[1] = tally[1] + 1
break
for y in range(len(wellbeing)):
for z in wellbeing[y]:
if x == z:
tally[2] = tally[2] + 1
break
# Handle a return greeting, and maybe ask how the user is
if tally[0] > 0:
sentence.append(greetings[randint(0, 6)])
if randint(0, 1) is 1:
sentence.append("how")
sentence.append("are")
sentence.append("you")
value = randint(0, 2)
if value is 0:
sentence.append("doing")
elif value is 1:
sentence.append("feeling")
# Handle questions about it's name with a simple answer
if tally[1] > 2 and len(input) < 5:
if randint(0, 1) is 1:
sentence.append("I")
sentence.append("am")
else:
sentence.append("my")
sentence.append("name")
sentence.append("is")
sentence.append(self.name)
# Handle a 'how are you' type question with a pre-determined emotional state
if tally[2] > 2 and len(input) < 5:
sentence.append("I")
sentence.append("am")
if randint(0, 1) is 1:
sentence.append("feeling")
sentence.append(self.feel)
return sentence
import time
import tensorflow as tf
import tensorlayer as tl
from sklearn.utils import shuffle
from tensorlayer.layers import EmbeddingInputlayer, Seq2Seq, DenseLayer, retrieve_seq_length_op2
from data.twitter import data
metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/')
(trainX, trainY), (testX, testY), (validX,
validY) = data.split_dataset(idx_q, idx_a)
trainX = trainX.tolist()
trainY = trainY.tolist()
testX = testX.tolist()
testY = testY.tolist()
validX = validX.tolist()
validY = validY.tolist()
trainX = tl.prepro.remove_pad_sequences(trainX)
trainY = tl.prepro.remove_pad_sequences(trainY)
testX = tl.prepro.remove_pad_sequences(testX)
testY = tl.prepro.remove_pad_sequences(testY)
validX = tl.prepro.remove_pad_sequences(validX)
validY = tl.prepro.remove_pad_sequences(validY)
# Hyperparameters
batch_size = 32
embedding_dimension = 1024
learning_rate = 0.0001
number_epochs = 1000
"""
import time
import tensorflow as tf
import tensorlayer as tl
import numpy as np
from tensorlayer.layers import DenseLayer, EmbeddingInputlayer, Seq2Seq, retrieve_seq_length_op2
from sklearn.utils import shuffle
from data.twitter import data
# Data Preparation
data_corpus = 'twitter'
metadata, idx_q, idx_a = data.load_data(PATH='data/{}/'.format(data_corpus))
(trainX, trainY), (testX, testY), (validX,
validY) = data.split_dataset(idx_q, idx_a)
trainX = trainX.tolist()
trainY = trainY.tolist()
testX = testX.tolist()
testY = testY.tolist()
validX = validX.tolist()
validY = validY.tolist()
trainX = tl.prepro.remove_pad_sequences(trainX)
trainY = tl.prepro.remove_pad_sequences(trainY)
testX = tl.prepro.remove_pad_sequences(testX)
testY = tl.prepro.remove_pad_sequences(testY)
validX = tl.prepro.remove_pad_sequences(validX)
References ---------- http://suriyadeepan.github.io/2016-12-31-practical-seq2seq/ """ import tensorflow as tf import tensorlayer as tl from tensorlayer.layers import * import tensorflow as tf import numpy as np import time ###============= prepare data from data.twitter import data metadata, idx_q, idx_a = data.load_data(PATH='seq2seq-chatbot/data/fb_chat/') # Twitter # from data.cornell_corpus import data # metadata, idx_q, idx_a = data.load_data(PATH='data/cornell_corpus/') # Cornell Moive (trainX, trainY), (testX, testY), (validX, validY) = data.split_dataset(idx_q, idx_a) trainX = trainX.tolist() trainY = trainY.tolist() testX = testX.tolist() testY = testY.tolist() validX = validX.tolist() validY = validY.tolist() trainX = tl.prepro.remove_pad_sequences(trainX) trainY = tl.prepro.remove_pad_sequences(trainY) testX = tl.prepro.remove_pad_sequences(testX) testY = tl.prepro.remove_pad_sequences(testY)
def initial_setup():
metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/')
return metadata
decode_ans = tl.prepro.pad_sequences(decode_ans)
mask = tl.prepro.sequences_get_mask(label_ans)
_, error = session.run(
[self.optimized_rnn, self.loss], {
encode_q: ques,
decode_a: decode_ans,
label_a: label_ans,
label_mask: mask
})
self.total_error += error
tl.files.save_npz(test_rnn.all_params, 'ChatbotRNN.npz', session)
if __name__ == "__main__":
frequentWords, questions, answers = data.load_data(PATH='data/twitter/')
word2id = frequentWords['w2idx']
id2word = frequentWords['idx2w']
word2id.update({'<GO>': len(id2word)})
word2id.update({'<EOS>': len(id2word) + 1})
size_of_dict = len(id2word) + 2
QuesTrain, AnsTrain, QuesTest, AnsTest = data.split_dataset(
questions, answers)
removeTrail = removeZeros()
assert removeTrail == True
QTrain, ATrain = shuffle(QTrain, ATrain, random_state=0)
encode_q = tf.placeholder(tf.int64, [cluster_size, None])
decode_a = tf.placeholder(tf.int64, [cluster_size, None])
def load_data(self): metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/') return metadata
def main():
metadata, idx_q, idx_a = data.load_data(PATH='data/twitter/')
trainX, trainY, testX, testY, validX, validY = getDataset(idx_q, idx_a)
xseq_len = len(trainX)
yseq_len = len(trainY)
assert xseq_len == yseq_len
batch_size = 32
n_step = int(xseq_len / batch_size)
xvocab_size = len(metadata['idx2w'])
emb_dim = 1024
w2idx = metadata['w2idx']
idx2w = metadata['idx2w']
unk_id = w2idx['unk']
pad_id = w2idx['_']
start_id = xvocab_size
end_id = xvocab_size + 1
w2idx.update({'start_id': start_id})
w2idx.update({'end_id': end_id})
idx2w = idx2w + ['start_id', 'end_id']
xvocab_size = yvocab_size = xvocab_size + 2
target_seqs = tl.prepro.sequences_add_end_id([trainY[10]],
end_id=end_id)[0]
decode_seqs = tl.prepro.sequences_add_start_id([trainY[10]],
start_id=start_id,
remove_last=False)[0]
target_mask = tl.prepro.sequences_get_mask([target_seqs])[0]
encode_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="encode_seqs")
decode_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="decode_seqs")
target_seqs = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="target_seqs")
target_mask = tf.placeholder(dtype=tf.int64,
shape=[batch_size, None],
name="target_mask")
net_out, _ = model(encode_seqs,
decode_seqs,
xvocab_size,
is_train=True,
reuse=False)
encode_seqs2 = tf.placeholder(dtype=tf.int64,
shape=[1, None],
name="encode_seqs")
decode_seqs2 = tf.placeholder(dtype=tf.int64,
shape=[1, None],
name="decode_seqs")
net, net_rnn = model(encode_seqs2,
decode_seqs2,
xvocab_size,
is_train=False,
reuse=True)
y = tf.nn.softmax(net.outputs)
loss = tl.cost.cross_entropy_seq_with_mask(logits=net_out.outputs,
target_seqs=target_seqs,
input_mask=target_mask,
return_details=False,
name='cost')
net_out.print_params(False)
lr = 0.0001
train_op = tf.train.AdamOptimizer(learning_rate=lr).minimize(loss)
gpu_option = tf.GPUOptions(per_process_gpu_memory_fraction=0.3)
sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False,
gpu_options=gpu_option))
tl.layers.initialize_global_variables(sess)
load_parameter = tl.files.load_and_assign_npz(sess=sess,
name='twitter.npz',
network=net)
if not load_parameter:
print("Loading npz fail, starting to train.")
n_epoch = 50
for epoch in range(n_epoch):
epoch_time = time.time()
from sklearn.utils import shuffle
trainX, trainY = shuffle(trainX, trainY, random_state=0)
total_err, n_iter = 0, 0
for X, Y in tl.iterate.minibatches(inputs=trainX,
targets=trainY,
batch_size=batch_size,
shuffle=False):
step_time = time.time()
X = tl.prepro.pad_sequences(X)
_target_seqs = tl.prepro.sequences_add_end_id(Y, end_id=end_id)
_target_seqs = tl.prepro.pad_sequences(_target_seqs)
_decode_seqs = tl.prepro.sequences_add_start_id(
Y, start_id=start_id, remove_last=False)
_decode_seqs = tl.prepro.pad_sequences(_decode_seqs)
_target_mask = tl.prepro.sequences_get_mask(_target_seqs)
_, err = sess.run(
[train_op, loss], {
encode_seqs: X,
decode_seqs: _decode_seqs,
target_seqs: _target_seqs,
target_mask: _target_mask
})
if n_iter % 200 == 0:
print("Epoch[%d/%d] step:[%d/%d] loss:%f took:%.5fs" %
(epoch, n_epoch, n_iter, n_step, err,
time.time() - step_time))
total_err += err
n_iter += 1
if n_iter % 1000 == 0:
print("Query> happy birthday to you")
getReplying(y, w2idx, idx2w, decode_seqs2, encode_seqs2,
start_id, end_id, sess, net_rnn,
"happy birthday to you")
print("Query> help me to do the exam")
getReplying(y, w2idx, idx2w, decode_seqs2, encode_seqs2,
start_id, end_id, sess, net_rnn,
"help me to do the exam")
print("Query> ny is so cold now")
getReplying(y, w2idx, idx2w, decode_seqs2, encode_seqs2,
start_id, end_id, sess, net_rnn,
"ny is so cold now")
print(
"Epoch[%d/%d] averaged loss:%f took:%.5fs" %
(epoch, n_epoch, total_err / n_iter, time.time() - epoch_time))
tl.files.save_npz(net.all_params, name='n.npz', sess=sess)
while (True):
getReplying(y, w2idx, idx2w, decode_seqs2, encode_seqs2, start_id,
end_id, sess, net_rnn, input("You>"))
