TensorFlow教程使用RNN生成唐诗

本教程转载至：TensorFlow练习7: 基于RNN生成古诗词

使用的数据集是全唐诗，首先提供一下数据集的下载链接：https://pan.baidu.com/s/13pNWfffr5HSN79WNb3Y0_w              提取码：koss

RNN不像传统的神经网络-它们的输出输出是固定的，而RNN允许我们输入输出向量序列。RNN是为了对序列数据进行建模而产生的。本帖代码移植自char-rnn，它是基于Torch的洋文模型，稍加修改即可应用于中文。char-rnn使用文本文件做为输入、训练RNN模型，然后使用它生成和训练数据类似的文本。

下边代码有修改，以适应TensorFlow1.4和GPU平台

 #coding=utf-8
 import collections
 import numpy as np
 import tensorflow as tf
 import io
 import sys
 import os
 reload(sys)
 sys.setdefaultencoding('utf-8')
 #-------------------------------数据预处理---------------------------#

 poetry_file ='poetry.txt'

 # 诗集
 poetrys = []
 with io.open(poetry_file, "r", encoding='utf-8',) as f:
     for line in f:
         # print line
         try:
             title, content = line.strip().split(':')
             content = content.replace(' ','')
             if '_' in content or '(' in content or '（' in content or '《' in content or '[' in content:
                 continue
             if len(content) < 5 or len(content) > 79:
                 continue
             content = '[' + content + ']'
             poetrys.append(content)
         except Exception as e:
             pass

 #按诗的字数排序
 poetrys = sorted(poetrys,key=lambda line: len(line))
 print(u"唐诗总数: ")
 print(len(poetrys))
 print(u"测试")

 # 统计每个字出现次数
 all_words = []
 for poetry in poetrys:
     all_words += [word for word in poetry]
 counter = collections.Counter(all_words)
 count_pairs = sorted(counter.items(), key=lambda x: -x[1])
 words, _ = zip(*count_pairs)

 # 取前多少个常用字
 words = words[:len(words)] + (' ',)
 # 每个字映射为一个数字ID
 word_num_map = dict(zip(words, range(len(words))))
 # 把诗转换为向量形式，参考TensorFlow练习1
 to_num = lambda word: word_num_map.get(word, len(words))
 poetrys_vector = [ list(map(to_num, poetry)) for poetry in poetrys]
 #[[314, 3199, 367, 1556, 26, 179, 680, 0, 3199, 41, 506, 40, 151, 4, 98, 1],
 #[339, 3, 133, 31, 302, 653, 512, 0, 37, 148, 294, 25, 54, 833, 3, 1, 965, 1315, 377, 1700, 562, 21, 37, 0, 2, 1253, 21, 36, 264, 877, 809, 1]
 #....]

 # 每次取64首诗进行训练
 batch_size = 64
 n_chunk = len(poetrys_vector) // batch_size
 x_batches = []
 y_batches = []
 for i in range(n_chunk):
     start_index = i * batch_size
     end_index = start_index + batch_size

     batches = poetrys_vector[start_index:end_index]
     length = max(map(len,batches))
     xdata = np.full((batch_size,length), word_num_map[' '], np.int32)
     for row in range(batch_size):
         xdata[row,:len(batches[row])] = batches[row]
     ydata = np.copy(xdata)
     ydata[:,:-1] = xdata[:,1:]
     """
     xdata             ydata
     [6,2,4,6,9]       [2,4,6,9,9]
     [1,4,2,8,5]       [4,2,8,5,5]
     """
     x_batches.append(xdata)
     y_batches.append(ydata)

 #---------------------------------------RNN--------------------------------------#

 input_data = tf.placeholder(tf.int32, [batch_size, None])
 output_targets = tf.placeholder(tf.int32, [batch_size, None])
 # 定义RNN
 def neural_network(model='lstm', rnn_size=128, num_layers=2):
     if model == 'rnn':
         cell_fun = tf.nn.rnn_cell.BasicRNNCell
     elif model == 'gru':
         cell_fun = tf.nn.rnn_cell.GRUCell
     elif model == 'lstm':
         cell_fun = tf.nn.rnn_cell.BasicLSTMCell

     cell = cell_fun(rnn_size, state_is_tuple=True)
     cell = tf.nn.rnn_cell.MultiRNNCell([cell] * num_layers, state_is_tuple=True)

     initial_state = cell.zero_state(batch_size, tf.float32)

     with tf.variable_scope('rnnlm'):
         softmax_w = tf.get_variable("softmax_w", [rnn_size, len(words)+1])
         softmax_b = tf.get_variable("softmax_b", [len(words)+1])
         with tf.device("/gpu:0"):
             embedding = tf.get_variable("embedding", [len(words)+1, rnn_size])
             inputs = tf.nn.embedding_lookup(embedding, input_data)

     outputs, last_state = tf.nn.dynamic_rnn(cell, inputs, initial_state=initial_state, scope='rnnlm')
     output = tf.reshape(outputs,[-1, rnn_size])

     logits = tf.matmul(output, softmax_w) + softmax_b
     probs = tf.nn.softmax(logits)
     return logits, last_state, probs, cell, initial_state

 ckpt_dir="./ckpt_dir"
 if not os.path.exists(ckpt_dir):
     os.makedirs(ckpt_dir)

 #训练
 def train_neural_network():
     logits, last_state, _, _, _ = neural_network()
     targets = tf.reshape(output_targets, [-1])
     loss = tf.contrib.legacy_seq2seq.sequence_loss_by_example([logits], [targets], [tf.ones_like(targets, dtype=tf.float32)], len(words))
     cost = tf.reduce_mean(loss)
     learning_rate = tf.Variable(0.0, trainable=False)
     tvars = tf.trainable_variables()
     grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars), 5)
     optimizer = tf.train.AdamOptimizer(learning_rate)
     train_op = optimizer.apply_gradients(zip(grads, tvars))

     with tf.Session() as sess:
         sess.run(tf.initialize_all_variables())

         saver = tf.train.Saver(tf.all_variables())

         for epoch in range(295):
             sess.run(tf.assign(learning_rate, 0.002 * (0.97 ** epoch)))
             n = 0
             for batche in range(n_chunk):
                 train_loss, _ , _ = sess.run([cost, last_state, train_op], feed_dict={input_data: x_batches[n], output_targets: y_batches[n]})
                 n += 1
                 print(epoch, batche, train_loss)
             if epoch % 7 == 0:
                 saver.save(sess, ckpt_dir+'/poetry.module', global_step=epoch)

 train_neural_network()

这里我只说自己对bug调试和调优的一些想法，具体代码理解，请联系作者本人。

首先是#coding=utf-8的问题，这里是告诉python环境，当前python脚本的文字编码是utf-8,这里如果不调整的话，默认的ansii环境极有可能报告编码错误。

之后是数据集的utf-8编码问题，这里在encoding的时候，用了utf-8 的选项，但是却没有告诉python环境，字符集编码是utf-8，会导致每次解析到的content和title都会报错，最终处理完的数据集大小为0，设置sys的默认编码可以解决。

同时，默认的open函数没有encoding选项，这个是在io.open中的选项，这个地方需要修改。

还有一点是一些接口使用问题，比如saver.save现在需要一个parent directory

之后是预测的代码

 #coding=utf-8
 import collections
 import numpy as np
 import tensorflow as tf
 import io
 import sys
 import os
 import pdb
 import time
 reload(sys)
 sys.setdefaultencoding('utf-8')
 #-------------------------------数据预处理---------------------------#

 poetry_file ='poetry.txt'

 # 诗集
 poetrys = []
 with io.open(poetry_file, "r", encoding='utf-8',) as f:
     for line in f:
         try:
             title, content = line.strip().split(':')
             content = content.replace(' ','')
             if '_' in content or '(' in content or '（' in content or '《' in content or '[' in content:
                 continue
             if len(content) < 5 or len(content) > 79:
                 continue
             content = '[' + content + ']'
             poetrys.append(content)
         except Exception as e:
             pass

 # 按诗的字数排序
 poetrys = sorted(poetrys,key=lambda line: len(line))
 print(u'唐诗总数: ', len(poetrys))

 # 统计每个字出现次数
 all_words = []
 for poetry in poetrys:
     all_words += [word for word in poetry]
 counter = collections.Counter(all_words)
 count_pairs = sorted(counter.items(), key=lambda x: -x[1])
 words, _ = zip(*count_pairs)

 # 取前多少个常用字
 words = words[:len(words)] + (' ',)
 # 每个字映射为一个数字ID
 word_num_map = dict(zip(words, range(len(words))))
 # 把诗转换为向量形式
 to_num = lambda word: word_num_map.get(word, len(words))
 poetrys_vector = [ list(map(to_num, poetry)) for poetry in poetrys]
 #[[314, 3199, 367, 1556, 26, 179, 680, 0, 3199, 41, 506, 40, 151, 4, 98, 1],
 #[339, 3, 133, 31, 302, 653, 512, 0, 37, 148, 294, 25, 54, 833, 3, 1, 965, 1315, 377, 1700, 562, 21, 37, 0, 2, 1253, 21, 36, 264, 877, 809, 1]
 #....]

 batch_size = 1
 n_chunk = len(poetrys_vector) // batch_size
 x_batches = []
 y_batches = []
 for i in range(n_chunk):
     start_index = i * batch_size
     end_index = start_index + batch_size

     batches = poetrys_vector[start_index:end_index]
     length = max(map(len,batches))
     xdata = np.full((batch_size,length), word_num_map[' '], np.int32)
     for row in range(batch_size):
         xdata[row,:len(batches[row])] = batches[row]
     ydata = np.copy(xdata)
     ydata[:,:-1] = xdata[:,1:]
     """
     xdata             ydata
     [6,2,4,6,9]       [2,4,6,9,9]
     [1,4,2,8,5]       [4,2,8,5,5]
     """
     x_batches.append(xdata)
     y_batches.append(ydata)

 #---------------------------------------RNN--------------------------------------#

 input_data = tf.placeholder(tf.int32, [batch_size, None])
 output_targets = tf.placeholder(tf.int32, [batch_size, None])
 # 定义RNN
 def neural_network(model='lstm', rnn_size=128, num_layers=2):
     if model == 'rnn':
         cell_fun = tf.nn.rnn_cell.BasicRNNCell
     elif model == 'gru':
         cell_fun = tf.nn.rnn_cell.GRUCell
     elif model == 'lstm':
         cell_fun = tf.nn.rnn_cell.BasicLSTMCell

     cell = cell_fun(rnn_size, state_is_tuple=True)
     cell = tf.nn.rnn_cell.MultiRNNCell([cell] * num_layers, state_is_tuple=True)

     initial_state = cell.zero_state(batch_size, tf.float32)

     with tf.variable_scope('rnnlm'):
         softmax_w = tf.get_variable("softmax_w", [rnn_size, len(words)+1])
         softmax_b = tf.get_variable("softmax_b", [len(words)+1])
         with tf.device("/gpu:0"):
             embedding = tf.get_variable("embedding", [len(words)+1, rnn_size])
             inputs = tf.nn.embedding_lookup(embedding, input_data)

     outputs, last_state = tf.nn.dynamic_rnn(cell, inputs, initial_state=initial_state, scope='rnnlm')
     output = tf.reshape(outputs,[-1, rnn_size])

     logits = tf.matmul(output, softmax_w) + softmax_b
     probs = tf.nn.softmax(logits)
     return logits, last_state, probs, cell, initial_state

 #-------------------------------生成古诗---------------------------------#
 # 使用训练完成的模型

 def gen_poetry():
     def to_word(weights):
         t = np.cumsum(weights)
         s = np.sum(weights)
         sample = int(np.searchsorted(t, np.random.rand(1)*s))
         return words[sample]

     _, last_state, probs, cell, initial_state = neural_network()

     with tf.Session() as sess:
         sess.run(tf.initialize_all_variables())

         saver = tf.train.Saver(tf.all_variables())
         saver.restore(sess, './ckpt_dir/poetry.module-294')

         state_ = sess.run(cell.zero_state(1, tf.float32))

         x = np.array([list(map(word_num_map.get, '['))])
         [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x, initial_state: state_})
         word = to_word(probs_)
         #word = words[np.argmax(probs_)]
         poem = ''
         while word != ']':
             poem += word
             x = np.zeros((1,1))
             x[0,0] = word_num_map[word]
             [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x, initial_state: state_})
             word = to_word(probs_)
             #word = words[np.argmax(probs_)]
         return poem

 def gen_poetry_with_head(head):
     def to_word(weights):
         t = np.cumsum(weights)
         s = np.sum(weights)
         sample = int(np.searchsorted(t, np.random.rand(1)*s))
         return words[sample]

     _, last_state, probs, cell, initial_state = neural_network()

     with tf.Session() as sess:
         sess.run(tf.initialize_all_variables())

         saver = tf.train.Saver(tf.all_variables())
         saver.restore(sess, './ckpt_dir/poetry.module-294')

         state_ = sess.run(cell.zero_state(1, tf.float32))
         poem = ''
         i = 0
         # print head
         # pdb.set_trace()
         for word in head:
             while word != '，' and word != '。':
                 poem += word
                 # print poem
                 # print head
                 # print word
                 x = np.array([list(map(word_num_map.get, word))])
                 [probs_, state_] = sess.run([probs, last_state], feed_dict={input_data: x, initial_state: state_})
                 word = to_word(probs_)
                 time.sleep(1)
             if i % 2 == 0:
                 poem += '，'
             else:
                 poem += '。'
             i += 1
         return poem

 print(gen_poetry())
 # print(gen_poetry_with_head(u'一二三四'))

这个藏头诗的代码用法有问题，不建议使用，我调了很久才调好，这次还是先列原作者的代码，下次单独说这块的调整和调优问题。

结果：

有那么点意思，但仔细看问题还是很大，胡言乱语，模型的调优远远不行。