import os
import numpy as np
import tensorflow as tf
def load_data(file_path):
"""
加载原始数据
:param file_path:
:return:
"""
with open(file_path, 'r') as reader:
data = reader.readlines()
return data
def create_lookup_table(text):
"""
构建字典表 {单词：数字} {数字：单词}
:param text:
:return:
"""
words = sorted(list(set(text)))
word2int = {word:idx for idx,word in enumerate(words)}
int2word = dict(enumerate(words))
return word2int, int2word
def create_X_and_Y(data, word2int, number_time_steps=3):
"""
基于原始数据，构建训练数据集的 X和Y
:param data:
:param word2int:
:param number_time_steps:
:return:
"""
X, Y = [], []
for content in data:
words = content.strip().split(' ')
words_number = len(words)
offset = 0
while offset < words_number - number_time_steps:
temp_x = words[offset: offset+number_time_steps]
temp_y = words[offset+number_time_steps]
X.append([word2int[tx] for tx in temp_x])
Y.append(word2int[temp_y])
offset +=1
X = np.asarray(X).reshape([-1, number_time_steps])
Y = np.asarray(Y).reshape(-1)
return X, Y
def create_model(vocab_size, num_units=32, number_time_steps=3):
"""
:param vocab_size: 词表大小
:param num_units: 隐藏层的节点数量（神经元个数）
:param number_time_steps: 时间步
:return:
"""
with tf.variable_scope('Network', initializer=tf.truncated_normal_initializer(stddev=0.1)):
with tf.variable_scope('input'):
"""
x:
[[2, 3, 4],
[7, 8, 9]]
y:
[[5],
[10]],
"""
_x = tf.placeholder(tf.int32, shape=[None, number_time_steps], name='x')
_y = tf.placeholder(tf.int32, shape=[None], name='y')
_x = tf.cast(_x, tf.float32)
input_x = tf.split(_x, num_or_size_splits=number_time_steps, axis=1)
with tf.variable_scope('rnn'):
cell_fw = tf.nn.rnn_cell.BasicLSTMCell(num_units=num_units)
cell_bw = tf.nn.rnn_cell.BasicLSTMCell(num_units=num_units)
rnn_outputs, _, _ = tf.nn.static_bidirectional_rnn(
cell_fw=cell_fw, cell_bw=cell_bw, inputs=input_x, dtype=tf.float32
)
with tf.variable_scope('logits'):
rnn_output = rnn_outputs[-1]
softmax_w = tf.get_variable(
'w', shape=[2*num_units, vocab_size], dtype=tf.float32
)
softmax_b = tf.get_variable(
'b', shape=[vocab_size], dtype=tf.float32, initializer=tf.zeros_initializer()
)
logits = tf.nn.xw_plus_b(rnn_output, softmax_w, softmax_b)
with tf.variable_scope('Predict'):
predictions = tf.argmax(logits, axis=1)
return _x, _y, logits, predictions
def create_loss(logits, labels):
"""
创建损失
:param logits:
:param labels:
:return:
"""
with tf.name_scope('loss'):
labels = tf.reshape(labels, shape=[-1])
loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=logits, labels=labels
))
return loss
def create_optimizer(loss, lr=1e-3):
"""
构建优化器
:param loss:
:param lr:
:return:
"""
with tf.name_scope('optimizer'):
optimizer = tf.train.AdamOptimizer(learning_rate=lr)
train_opt = optimizer.minimize(loss)
return train_opt
def train(checkpoint_dir, max_steps=10000, batch_size=64, num_units=32, number_time_steps=10):
graph = tf.Graph()
with graph.as_default():
data = load_data(file_path='../datas/belling_the_cat.txt')
text = []
for line in data:
line = line.strip()
for word in line.split(' '):
text.append(word)
word2int, int2word = create_lookup_table(text)
x, y = create_X_and_Y(data, word2int, number_time_steps=number_time_steps)
_x, _y, logits, predictions = create_model(
len(word2int), num_units=num_units, number_time_steps=number_time_steps)
loss = create_loss(logits, _y)
train_opt = create_optimizer(loss)
saver = tf.train.Saver()
with tf.Session(graph=graph) as sess:
sess.run(tf.global_variables_initializer())
total_samples = x.shape[0]
n_batches = total_samples // batch_size
time = 0
random_index = np.random.permutation(total_samples)
for step in range(1, max_steps):
start_idx = time * batch_size
end_idx = start_idx + batch_size
idx = random_index[start_idx: end_idx]
train_x = x[idx]
train_y = y[idx]
feed = {_x: train_x, _y: train_y}
sess.run(train_opt, feed)
if step % 200==0:
train_loss = sess.run(loss, feed)
print('step:{} - Train loss:{}'.format(step, train_loss))
index = np.random.randint(low=0, high=total_samples)
sample_in = np.reshape(x[index], newshape=[-1, number_time_steps])
sample_out = sess.run(predictions, feed_dict={_x: sample_in})
print('输入:{} - 预测:{} VS 真实值:{}'.format(
x[index], int2word[sample_out[0]], int2word[y[index]]))
if step % 1000 == 0:
files = 'model.ckpt'
save_files = os.path.join(checkpoint_dir, files)
saver.save(sess, save_path=save_files, global_step=step)
print('model saved!!')
time += 1
if time == n_batches:
time =0
random_index = np.random.permutation(total_samples)
if __name__ == '__main__':
checkpoint_dir = './models'
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
train(
checkpoint_dir, max_steps=10000, batch_size=64, num_units=32, number_time_steps=10
)
D:\Anaconda\python.exe D:/AI20/HJZ/04-深度学习/4-RNN/20191228___AI20_RNN/03_双向rnn_英文小短文.py
2020-02-18 10:42:51.076290: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
step:200 - Train loss:3.9305477142333984
输入:[ 8 7 1 76 39 1 85 66 24 30] - 预测:, VS 真实值:consists
step:400 - Train loss:3.4807467460632324
输入:[ 77 100 44 86 60 2 91 69 56 109] - 预测:, VS 真实值:general
step:600 - Train loss:2.976116418838501
输入:[ 86 59 63 86 23 2 21 91 53 101] - 预测:the VS 真实值:should
step:800 - Train loss:2.6706745624542236
输入:[65 12 11 61 83 2 88 86 64 58] - 预测:the VS 真实值:said
step:1000 - Train loss:2.3042399883270264
输入:[ 39 92 110 55 86 22 2 111 108 8] - 预测:agree VS 真实值:agree
model saved!!
step:1200 - Train loss:1.834552526473999
输入:[64 58 37 96 11 76 0 85 45 8] - 预测:very VS 真实值:very
step:1400 - Train loss:1.6621453762054443
输入:[ 40 14 1 101 28 31 34 35 40 2] - 预测:i VS 真实值:i
step:1600 - Train loss:1.3067286014556885
输入:[ 29 93 26 103 54 90 28 84 93 67] - 预测:their VS 真实值:their
step:1800 - Train loss:1.2478928565979004
输入:[ 1 86 57 38 4 36 29 93 26 103] - 预测:measures VS 真实值:measures
step:2000 - Train loss:0.9508646726608276
输入:[ 92 110 55 86 22 2 111 108 8 7] - 预测:, VS 真实值:,
model saved!!
step:2200 - Train loss:0.8384971618652344
输入:[61 83 2 88 86 64 58 76 46 45] - 预测:easy VS 真实值:easy
step:2400 - Train loss:0.5928971171379089
输入:[64 58 37 96 11 76 0 85 45 8] - 预测:very VS 真实值:very
step:2600 - Train loss:0.5592891573905945
输入:[31 34 35 40 2 41 98 1 89 1] - 预测:to VS 真实值:to
step:2800 - Train loss:0.4942898452281952
输入:[ 14 1 101 28 31 34 35 40 2 41] - 预测:venture VS 真实值:venture
step:3000 - Train loss:0.35549503564834595
输入:[ 77 100 44 86 60 2 91 69 56 109] - 预测:general VS 真实值:general
model saved!!
step:3200 - Train loss:0.3749193549156189
输入:[93 19 86 23 3 86 57 50 16 65] - 预测:another VS 真实值:another
step:3400 - Train loss:0.29113298654556274
输入:[86 23 2 82 76 91 1 11 82 76] - 预测:that VS 真实值:that
step:3600 - Train loss:0.25510722398757935
输入:[93 70 85 4 81 19 18 68 1 11] - 预测:attached VS 真实值:attached
step:3800 - Train loss:0.2347201555967331
输入:[ 26 103 54 90 28 84 93 67 87 25] - 预测:enemy VS 真实值:enemy
step:4000 - Train loss:0.18923482298851013
输入:[ 1 95 10 64 58 37 96 11 76 0] - 预测:that VS 真实值:that
model saved!!
step:4200 - Train loss:0.16195067763328552
输入:[23 3 86 57 50 16 65 12 11 61] - 预测:spoke VS 真实值:spoke
step:4400 - Train loss:0.13978148996829987
输入:[ 86 59 63 86 23 2 21 91 53 101] - 预测:should VS 真实值:should
step:4600 - Train loss:0.1490730196237564
输入:[103 54 90 28 84 93 67 87 25 33] - 预测:, VS 真实值:,
step:4800 - Train loss:0.10961226373910904
输入:[ 1 11 82 76 85 20 16 48 4 112] - 预测:mouse VS 真实值:mouse
step:5000 - Train loss:0.11105622351169586
输入:[ 86 22 2 111 108 8 7 1 76 39] - 预测:, VS 真实值:,
model saved!!
step:5200 - Train loss:0.0975622832775116
输入:[ 78 9 47 104 77 100 5 1 11 28] - 预测:easily VS 真实值:easily
step:5400 - Train loss:0.0716937854886055
输入:[18 68 1 11 17 21 4 74 75 86] - 预测:neck VS 真实值:neck
step:5600 - Train loss:0.07302534580230713
输入:[ 91 53 101 78 9 47 104 77 100 5] - 预测:, VS 真实值:,
step:5800 - Train loss:0.05743904039263725
输入:[66 24 30 27 44 86 80 11 94 52] - 预测:in VS 真实值:in
step:6000 - Train loss:0.05397602543234825
输入:[ 1 85 66 24 30 27 44 86 80 11] - 预测:treacherous VS 真实值:treacherous
model saved!!
step:6200 - Train loss:0.054213933646678925
输入:[87 25 33 1 86 23 2 82 76 91] - 预测:, VS 真实值:,
step:6400 - Train loss:0.0373719185590744
输入:[23 2 82 76 91 1 11 82 76 85] - 预测:but VS 真实值:but
step:6600 - Train loss:0.046218693256378174
输入:[ 42 101 28 71 82 79 63 40 14 1] - 预测:we VS 真实值:we
step:6800 - Train loss:0.03185339272022247
输入:[ 20 16 48 4 112 58 37 96 11 76] - 预测:he VS 真实值:he
step:7000 - Train loss:0.026730481535196304
输入:[ 1 11 17 21 4 74 75 86 59 63] - 预测:the VS 真实值:the
model saved!!
step:7200 - Train loss:0.02903711423277855
输入:[111 108 8 7 1 76 39 1 85 66] - 预测:chief VS 真实值:chief
step:7400 - Train loss:0.026526065543293953
输入:[ 57 38 4 36 29 93 26 103 54 90] - 预测:could VS 真实值:could
step:7600 - Train loss:0.02054942026734352
输入:[107 45 93 19 86 23 3 86 57 50] - 预测:at VS 真实值:at
step:7800 - Train loss:0.01777688041329384
输入:[ 85 20 16 48 4 112 58 37 96 11] - 预测:said VS 真实值:said
step:8000 - Train loss:0.014596270397305489
输入:[112 58 37 96 11 76 39 38 4 69] - 预测:to VS 真实值:to
model saved!!
step:8200 - Train loss:0.015546170994639397
输入:[107 45 93 19 86 23 3 86 57 50] - 预测:at VS 真实值:at
step:8400 - Train loss:0.01338121946901083
输入:[16 65 12 11 61 83 2 88 86 64] - 预测:mouse VS 真实值:mouse
step:8600 - Train loss:0.014673653990030289
输入:[ 1 85 66 24 30 27 44 86 80 11] - 预测:treacherous VS 真实值:treacherous
step:8800 - Train loss:0.010602903552353382
输入:[75 86 59 63 86 23 2 21 91 53] - 预测:we VS 真实值:we
step:9000 - Train loss:0.01917443238198757
输入:[ 1 20 107 45 93 19 86 23 3 86] - 预测:mice VS 真实值:mice
model saved!!
step:9200 - Train loss:0.012528151273727417
输入:[ 1 89 1 93 70 85 4 81 19 18] - 预测:procured VS 真实值:procured
step:9400 - Train loss:0.009897150099277496
输入:[ 26 103 54 90 28 84 93 67 87 25] - 预测:enemy VS 真实值:enemy
step:9600 - Train loss:0.007559692487120628
输入:[ 20 16 48 4 112 58 37 96 11 76] - 预测:he VS 真实值:he
step:9800 - Train loss:0.008503235876560211
输入:[33 1 86 23 2 82 76 91 1 11] - 预测:some VS 真实值:some
Process finished with exit code 0
