tensorflow1.12.0 seq2seq基础练习

代码来自《唐宇迪seq2seq》,原本环境tf1.0,经过一些更改，tf1.12.0

训练目标：

将输入的单词按照字母的自然顺序进行排序
Word 编号:    [24, 17, 9, 9, 28, 0, 0]
Input Words: h e l l o <PAD> <PAD>
Word 编号:       [17, 24, 9, 9, 28, 3]
Response Words: e h l l o <EOS>

重要的点：
seq2seq主要由encoder网络和decoder网络组成
1.为了使得输入和输出规范，对每个单词进行规范，多添加了[’’, ‘’, ‘’, ‘’]4个特殊字符
''填充, ''特殊或不存在的字符串, ''起始, ''结束
2.对输入的字符都统一的映射到同一维度空间tf.contrib.layers.embed_sequence
3.其实encoder网络就是RNN
4.decoder网络接受encoder网络的RNN 输出state(如果RNN为basicRNN,则为RNN最后一step的输出，如果为LSTM，state为LSTM最后一step的输出，还有门信息)，然后又再进行多层RNN的解析，最后经过全连接层输出结果
5.decoder网络中还使用到了TrainingHelper帮助训练(当然也可以不适用，但使用效果更好)
tf.contrib.seq2seq.TrainingHelper这个函数不会把t-1阶段的输出作为t阶段的输入，而是把target中的真实值直接输入给RNN。

数据可以自己造,找一些英文文档，分词以后就形成source了，然后对分词后的词进行按字母排序就是target了，很简单

示例代码

# coding=utf-8
'''
Created on 2019年4月13日

@author: admin
'''
import numpy as np
import time
import tensorflow as tf
from tensorflow.python.layers.core import Dense
import os

# 超参数
# Number of Epochs
epochs = 60
# Batch Size
batch_size = 128
# RNN Size
rnn_size = 50
# Number of Layers
num_layers = 2
# Embedding Size
encoding_embedding_size = 15
decoding_embedding_size = 15
# Learning Rate
learning_rate = 0.001

with open('data/letters_source.txt', 'r', encoding='utf-8') as f:
    source_data = f.read()

with open('data/letters_target.txt', 'r', encoding='utf-8') as f:
    target_data = f.read()
    
def extract_character_vocab(data):
    '''
    构造映射表
    '''
    #'<PAD>'填充, '<UNK>'特殊或不存在的字符串, '<GO>'起始,  '<EOS>'结束
    special_words = ['<PAD>', '<UNK>', '<GO>',  '<EOS>']

    set_words = list(set([character for line in data.split('\n') for character in line]))
    # 这里要把四个特殊字符添加进词典
    int_to_vocab = {idx: word for idx, word in enumerate(special_words + set_words)}
    vocab_to_int = {word: idx for idx, word in int_to_vocab.items()}

    return int_to_vocab, vocab_to_int

# 构造映射表
source_int_to_letter, source_letter_to_int = extract_character_vocab(source_data)
target_int_to_letter, target_letter_to_int = extract_character_vocab(target_data)

# 对字母进行转换
source_int = [[source_letter_to_int.get(letter, source_letter_to_int['<UNK>']) 
               for letter in line] for line in source_data.split('\n')]
target_int = [[target_letter_to_int.get(letter, target_letter_to_int['<UNK>']) 
               for letter in line] + [target_letter_to_int['<EOS>']] for line in target_data.split('\n')] 

def get_inputs():
    '''
    模型输入tensor
    '''
    inputs = tf.placeholder(tf.int32, [None, None], name='inputs')
    targets = tf.placeholder(tf.int32, [None, None], name='targets')
    learning_rate = tf.placeholder(tf.float32, name='learning_rate')
       
    # 定义target序列最大长度（之后target_sequence_length和source_sequence_length会作为feed_dict的参数）
    target_sequence_length = tf.placeholder(tf.int32, (None,), name='target_sequence_length')
    max_target_sequence_length = tf.reduce_max(target_sequence_length, name='max_target_len')
    source_sequence_length = tf.placeholder(tf.int32, (None,), name='source_sequence_length')
       
    return inputs, targets, learning_rate, target_sequence_length, max_target_sequence_length, source_sequence_length
   
def get_encoder_layer(input_data, rnn_size, num_layers,
                   source_sequence_length, source_vocab_size, 
                   encoding_embedding_size):
   
    '''
    构造Encoder层
       
    参数说明：
    - input_data: 输入tensor
    - rnn_size: rnn隐层结点数量
    - num_layers: 堆叠的rnn cell数量
    - source_sequence_length: 源数据的序列长度
    - source_vocab_size: 源数据的词典大小
    - encoding_embedding_size: embedding的大小
    '''
    # Encoder embedding
    encoder_embed_input = tf.contrib.layers.embed_sequence(input_data, source_vocab_size, encoding_embedding_size)
   
    # RNN cell
    def get_lstm_cell(rnn_size):
        lstm_cell = tf.contrib.rnn.LSTMCell(rnn_size, initializer=tf.random_uniform_initializer(-0.1, 0.1, seed=2))
        return lstm_cell
   
    cell = tf.contrib.rnn.MultiRNNCell([get_lstm_cell(rnn_size) for _ in range(num_layers)])
       
    #dynamic_rnn与rnn的区别在于，dynamic_rnn对于不同的batch，可以接收不同的sequence_length。
    #例如，第一个batch是[batch_size,10]，第二个batch是[batch_size,20]。而rnn只能接收定长的sequence_length。
    encoder_output, encoder_state = tf.nn.dynamic_rnn(cell, encoder_embed_input, 
                                                      sequence_length=source_sequence_length, dtype=tf.float32)
       
    return encoder_output, encoder_state
   
def process_decoder_input(data, vocab_to_int, batch_size):
    '''
    补充<GO>，并移除最后一个字符
    '''
    # cut掉最后一个字符，然后将'<GO>'加到第一列
    ending = tf.strided_slice(data, [0, 0], [batch_size, -1], [1, 1])
    decoder_input = tf.concat([tf.fill([batch_size, 1], vocab_to_int['<GO>']), ending], 1)
   
    return decoder_input
   
   
def decoding_layer(target_letter_to_int, decoding_embedding_size, num_layers, rnn_size,
                   target_sequence_length, max_target_sequence_length, encoder_state, decoder_input):
    '''
    构造Decoder层
       
    参数：
    - target_letter_to_int: target数据的映射表
    - decoding_embedding_size: embed向量大小
    - num_layers: 堆叠的RNN单元数量
    - rnn_size: RNN单元的隐层结点数量
    - target_sequence_length: target数据序列长度
    - max_target_sequence_length: target数据序列最大长度
    - encoder_state: encoder端编码的状态向量
    - decoder_input: decoder端输入
    '''
    # 1. Embedding
    target_vocab_size = len(target_letter_to_int)
    decoder_embeddings = tf.Variable(tf.random_uniform([target_vocab_size, decoding_embedding_size]))
    decoder_embed_input = tf.nn.embedding_lookup(decoder_embeddings, decoder_input)
   
    # 2. 构造Decoder中的RNN单元
    def get_decoder_cell(rnn_size):
        decoder_cell = tf.contrib.rnn.LSTMCell(rnn_size,
                                           initializer=tf.random_uniform_initializer(-0.1, 0.1, seed=2))
        return decoder_cell
    cell = tf.contrib.rnn.MultiRNNCell([get_decoder_cell(rnn_size) for _ in range(num_layers)])
        
    # 3. Output全连接层
    output_layer = Dense(target_vocab_size,
                         kernel_initializer = tf.truncated_normal_initializer(mean = 0.0, stddev=0.1))
   
   
    # 4. Training decoder
    with tf.variable_scope("decode"):
        # 得到help对象，这个函数不会把t-1阶段的输出作为t阶段的输入，而是把target中的真实值直接输入给RNN。
        training_helper = tf.contrib.seq2seq.TrainingHelper(inputs=decoder_embed_input,
                                                            sequence_length=target_sequence_length,
                                                            time_major=False)
        # 构造decoder
        training_decoder = tf.contrib.seq2seq.BasicDecoder(cell,
                                                           training_helper,
                                                           encoder_state,
                                                           output_layer)
        # tf.contrib.seq2seq.dynamic_decode在低版本中只Return两个，这里1.12版本会返回3个
        training_decoder_output,_,_ = tf.contrib.seq2seq.dynamic_decode(training_decoder,
                                                                       impute_finished=True,
                                                                       maximum_iterations=max_target_sequence_length)
    # 5. Predicting decoder
    # 与training共享参数
    with tf.variable_scope("decode", reuse=True):
        # 创建一个常量tensor并复制为batch_size的大小
        start_tokens = tf.tile(tf.constant([target_letter_to_int['<GO>']], dtype=tf.int32), [batch_size], 
                               name='start_tokens')
        #它和TrainingHelper的区别在于它会把t-1下的输出进行embedding后再输入给RNN。
        predicting_helper = tf.contrib.seq2seq.GreedyEmbeddingHelper(decoder_embeddings,
                                                                start_tokens,
                                                                target_letter_to_int['<EOS>'])
        predicting_decoder = tf.contrib.seq2seq.BasicDecoder(cell,
                                                        predicting_helper,
                                                        encoder_state,
                                                        output_layer)
        predicting_decoder_output,_,_ = tf.contrib.seq2seq.dynamic_decode(predicting_decoder,
                                                            impute_finished=True,
                                                            maximum_iterations=max_target_sequence_length)
       
    return training_decoder_output, predicting_decoder_output
   
def seq2seq_model(input_data, targets, lr, target_sequence_length, 
                  max_target_sequence_length, source_sequence_length,
                  source_vocab_size, target_vocab_size,
                  encoder_embedding_size, decoder_embedding_size, 
                  rnn_size, num_layers):
       
    # 获取encoder的状态输出
    _, encoder_state = get_encoder_layer(input_data, 
                                  rnn_size, 
                                  num_layers, 
                                  source_sequence_length,
                                  source_vocab_size, 
                                  encoding_embedding_size)
       
    # 预处理后的decoder输入
    decoder_input = process_decoder_input(targets, target_letter_to_int, batch_size)
       
    # 将状态向量与输入传递给decoder
    training_decoder_output, predicting_decoder_output = decoding_layer(target_letter_to_int, 
                                                                       decoding_embedding_size, 
                                                                       num_layers, 
                                                                       rnn_size,
                                                                       target_sequence_length,
                                                                       max_target_sequence_length,
                                                                       encoder_state, 
                                                                       decoder_input) 
       
    return training_decoder_output, predicting_decoder_output
   
# # 构造graph
train_graph = tf.Graph()
    
with train_graph.as_default():
       
    # 获得模型输入    
    input_data, targets, lr, target_sequence_length, max_target_sequence_length, source_sequence_length = get_inputs()
       
    training_decoder_output, predicting_decoder_output = seq2seq_model(input_data, 
                                                                      targets, 
                                                                      lr, 
                                                                      target_sequence_length, 
                                                                      max_target_sequence_length, 
                                                                      source_sequence_length,
                                                                      len(source_letter_to_int),
                                                                      len(target_letter_to_int),
                                                                      encoding_embedding_size, 
                                                                      decoding_embedding_size, 
                                                                      rnn_size, 
                                                                      num_layers)    
       
    training_logits = tf.identity(training_decoder_output.rnn_output, 'logits')
    predicting_logits = tf.identity(predicting_decoder_output.sample_id, name='predictions')
    print(predicting_logits)
       
    masks = tf.sequence_mask(target_sequence_length, max_target_sequence_length, dtype=tf.float32, name='masks')
    
    with tf.name_scope("optimization"):
           
        # Loss function
        cost = tf.contrib.seq2seq.sequence_loss(
            training_logits,
            targets,
            masks)
        print(cost)
    
        # Optimizer
        optimizer = tf.train.AdamOptimizer(lr)
    
        # Gradient Clipping 基于定义的min与max对tesor数据进行截断操作，目的是为了应对梯度爆发或者梯度消失的情况
        gradients = optimizer.compute_gradients(cost)
        capped_gradients = [(tf.clip_by_value(grad, -5., 5.), var) for grad, var in gradients if grad is not None]
        train_op = optimizer.apply_gradients(capped_gradients)
   
def pad_sentence_batch(sentence_batch, pad_int):
    '''
    对batch中的序列进行补全，保证batch中的每行都有相同的sequence_length
       
    参数：
    - sentence batch
    - pad_int: <PAD>对应索引号
    '''
    max_sentence = max([len(sentence) for sentence in sentence_batch])
    return [sentence + [pad_int] * (max_sentence - len(sentence)) for sentence in sentence_batch]
   
def get_batches(targets, sources, batch_size, source_pad_int, target_pad_int):
    '''
    定义生成器，用来获取batch
    '''
    for batch_i in range(0, len(sources)//batch_size):
        start_i = batch_i * batch_size
        sources_batch = sources[start_i:start_i + batch_size]
        targets_batch = targets[start_i:start_i + batch_size]
        # 补全序列，batch中的每行都有相同的sequence_length
        pad_sources_batch = np.array(pad_sentence_batch(sources_batch, source_pad_int))
        pad_targets_batch = np.array(pad_sentence_batch(targets_batch, target_pad_int))
           
        # 记录每条记录的长度
        pad_targets_lengths = []
        for target in pad_targets_batch:
            pad_targets_lengths.append(len(target))
           
        pad_source_lengths = []
        for source in pad_sources_batch:
            pad_source_lengths.append(len(source))
           
        yield pad_targets_batch, pad_sources_batch, pad_targets_lengths, pad_source_lengths
   
# 将数据集分割为train和validation
train_source = source_int[batch_size:]
train_target = target_int[batch_size:]
# 留出一个batch进行验证
valid_source = source_int[:batch_size]
valid_target = target_int[:batch_size]
(valid_targets_batch, valid_sources_batch, valid_targets_lengths, valid_sources_lengths) = next(get_batches(valid_target, valid_source, batch_size,
                           source_letter_to_int['<PAD>'],
                           target_letter_to_int['<PAD>']))
   
display_step = 50 # 每隔50轮输出loss
# # 保存模型超参数
ckpt_dir = "seq2seq_ckpt"
if not os.path.exists(ckpt_dir):
    os.makedirs(ckpt_dir)

checkpoint = ckpt_dir+"/trained_model.ckpt" 
with tf.Session(graph=train_graph) as sess:
    sess.run(tf.global_variables_initializer())
        
    for epoch_i in range(1, epochs+1):
        for batch_i, (targets_batch, sources_batch, targets_lengths, sources_lengths) in enumerate(
                get_batches(train_target, train_source, batch_size,
                           source_letter_to_int['<PAD>'],
                           target_letter_to_int['<PAD>'])):
            
            _, loss = sess.run(
                [train_op, cost],
                {input_data: sources_batch,
                 targets: targets_batch,
                 lr: learning_rate,
                 target_sequence_length: targets_lengths,
                 source_sequence_length: sources_lengths})

            if batch_i % display_step == 0:
                
                # 计算validation loss
                validation_loss = sess.run(
                [cost],
                {input_data: valid_sources_batch,
                 targets: valid_targets_batch,
                 lr: learning_rate,
                 target_sequence_length: valid_targets_lengths,
                 source_sequence_length: valid_sources_lengths})
                
                print('Epoch {:>3}/{} Batch {:>4}/{} - Training Loss: {:>6.3f}  - Validation loss: {:>6.3f}'
                      .format(epoch_i,
                              epochs, 
                              batch_i, 
                              len(train_source) // batch_size, 
                              loss, 
                              validation_loss[0]))

    
    
    # 保存模型
    saver = tf.train.Saver()
    saver.save(sess, checkpoint)
    print('Model Trained and Saved')

#训练结束后开始测试   
def source_to_seq(text):
    '''
    对源数据进行转换
    '''
    sequence_length = 7
    return [source_letter_to_int.get(word, source_letter_to_int['<UNK>']) for word in text] + [source_letter_to_int['<PAD>']]*(sequence_length-len(text))


# 输入一个单词
input_word = 'hello'
text = source_to_seq(input_word)

loaded_graph = tf.Graph()
with tf.Session(graph=loaded_graph) as sess:
    # 加载模型
    loader = tf.train.import_meta_graph(checkpoint + '.meta')
    loader.restore(sess, checkpoint)

    input_data = loaded_graph.get_tensor_by_name('inputs:0')
    logits = loaded_graph.get_tensor_by_name('predictions:0')
    source_sequence_length = loaded_graph.get_tensor_by_name('source_sequence_length:0')
    target_sequence_length = loaded_graph.get_tensor_by_name('target_sequence_length:0')
    
    answer_logits = sess.run(logits, {input_data: [text]*batch_size, 
                                      target_sequence_length: [len(text)]*batch_size, 
                                      source_sequence_length: [len(text)]*batch_size})[0]
     


pad = source_letter_to_int["<PAD>"] 

print('原始输入:', input_word)

print('\nSource')
print('  Word 编号:    {}'.format([i for i in text]))
print('  Input Words: {}'.format(" ".join([source_int_to_letter[i] for i in text])))

# print('\nTarget')
print('  Word 编号:       {}'.format([i for i in answer_logits]))
print('  Response Words: {}'.format(" ".join([target_int_to_letter[i] for i in answer_logits])))