‘‘‘
Created on 2017年5月13日

@author: weizhen
‘‘‘
import numpy as np
import tensorflow as tf
import ptb_iterator as reader
from tensorflow.contrib import rnn 

DATA_PATH = "/path/to/ptb/data"  # 数据存放的路径
HIDDEN_SIZE = 200  # 隐藏层的规模
NUM_LAYERS = 2  # 深层循环神经网络中LSTM结构的层数
VOCAB_SIZE = 10000  # 词典规模，加上语句结束标识符和稀有单词标识符总共一万个单词

LEARNING_RATE = 1.0  # 学习速率
TRAIN_BATCH_SIZE = 20  # 训练数据batch的大小
TRAIN_NUM_STEP = 35  # 训练数据截断长度

# 在测试时不需要使用截断，所以可以将测试数据看成一个超长的序列
EVAL_BATCH_SIZE = 1  # 测试数据batch的大小
EVAL_NUM_STEP = 1  # 测试数据截断长度
NUM_EPOCH = 2  # 使用训练数据的轮数
KEEP_PROB = 0.5  # 节点不被dropout的概率
MAX_GRAD_NORM = 5  # 用于控制梯度膨胀的参数

def LstmCell(is_training):
    lstm_cell = rnn.BasicLSTMCell(HIDDEN_SIZE,reuse=tf.get_variable_scope().reuse)
    if is_training:
            lstm_cell = rnn.DropoutWrapper(lstm_cell, output_keep_prob=KEEP_PROB)
    return lstm_cell
# 通过一个PTBModel类来描述模型，这样方便维护循环神经网络中的状态
class PTBModel(object):
    def __init__(self, is_training, batch_size, num_steps):
        # 记录使用的batch大小和截断长度
        self.batch_size = batch_size
        self.num_steps = num_steps
        
        # 定义输入层。可以看到输入层的维度为batch_size*num_steps,这和
        # ptb_iterator函数输出的训练数据batch是一致的
        self.input_data =http://www.mamicode.com/ tf.placeholder(tf.int32, [batch_size, num_steps])
        
        # 定义预期输出，它的维度和ptb_iterator函数输出的正确答案维度也是一样的
        self.targets = tf.placeholder(tf.int32, [batch_size, num_steps])
        
        # 定义使用LSTM结构为循环体结构且使用dropout的深层循环神经网络
#         lstm_cell = rnn.BasicLSTMCell(HIDDEN_SIZE)
#         if is_training:
#             lstm_cell = rnn.DropoutWrapper(lstm_cell, output_keep_prob=KEEP_PROB)
        cell = rnn.MultiRNNCell([LstmCell(is_training) for _ in range(NUM_LAYERS)])
        
        # 初始化最初的状态，也就是全零的向量
        self.initial_state = cell.zero_state(batch_size, tf.float32)
        
        # 将单词ID转换成为单词向量。因为总共有VOCAB_SIZE个单词，每个单词向量的维度为HIDDEN_SIZE
        # 所以embedding参数的维度为VOCAB_SIZE*HIDDEN_SIZE
        embedding = tf.get_variable("embedding", [VOCAB_SIZE, HIDDEN_SIZE])
        
        # 将原本batch_size*num_steps个单词ID转化为单词向量，转换后的输入层维度为batch_size*num_steps*HIDDEN_SIZE
        inputs = tf.nn.embedding_lookup(embedding, self.input_data)
        
        # 只在训练时使用dropout
        if is_training:
            inputs = tf.nn.dropout(inputs, KEEP_PROB)
        
        # 定义输出列表，在这里先将不同时刻LSTM结构的输出收集起来，再通过一个全连接层得到最终的输出
        outputs = []
        # state存储不同batch种LSTM的状态，将其初始化为0
        state = self.initial_state
        with tf.variable_scope("RNN"):
            for time_step in range(num_steps):
                if time_step > 0: tf.get_variable_scope().reuse_variables()
                # 从输入数据中获取当前时刻的输入并传入LSTM结构
                
                cell_output, state = cell(inputs[:, time_step, :], state)
                #cell_output, state = tf.nn.dynamic_rnn(cell,inputs[:, time_step, :], state,time_major=False)
                # 当前输出加入输出队列
                outputs.append(cell_output)
        
        # 把输出队列展开成[batch*hidden_size*num_steps]的形状，然后再
        # reshape成[batch*numsteps,hidden_size]的形状
        output = tf.reshape(tf.concat(outputs, 1), [-1, HIDDEN_SIZE])
        
        # 将从LSTM中得到的输出再经过一个全连接层得到最后的预测结果，最终的预测结果在每一个时刻上都是一个长度为VOCAB_Size的数组，
        # 经过softmax层之后表示下一个位置是不同单词的概率
        weight = tf.get_variable("weight", [HIDDEN_SIZE, VOCAB_SIZE])
        bias = tf.get_variable("bias", [VOCAB_SIZE])
        logits = tf.matmul(output, weight) + bias
        
        # 定义交叉熵损失函数。TensorFlow提供了sequence_loss_by_example函数来计算一个序列的交叉熵的和
        loss = tf.contrib.legacy_seq2seq.sequence_loss_by_example(
            [logits],  # 预测的结果
            [tf.reshape(self.targets, [-1])],  # 期待的正确答案，这里讲[batch_size,num_steps]二维数组压缩成一维数组
            [tf.ones([batch_size * num_steps], dtype=tf.float32)]  # 损失的权重，在这里所有的权重都为1，也就是说不同batch和不同时刻的重要程度是一样的
            )
        
        # 计算得到每个batch的平均损失
        self.cost = tf.reduce_sum(loss) / batch_size
        self.final_state = state
        
        # 只在训练模型时定义方向传播操作
        if not is_training:
            return
        trainable_variables = tf.trainable_variables()
        # 通过clip_by_global_norm函数控制梯度的大小，避免梯度膨胀的问题
        grads, _ = tf.clip_by_global_norm(tf.gradients(self.cost, trainable_variables), MAX_GRAD_NORM)
        
        # 定义优化方法
        optimizer = tf.train.GradientDescentOptimizer(LEARNING_RATE)
        # 定义训练步骤
        self.train_op = optimizer.apply_gradients(zip(grads, trainable_variables))
        
# 使用给定的模型model在数据data上运行train_op并返回在全部数据上的perplexity值
def run_epoch(session, model, data, train_op, output_log):
    # 计算perplexity的辅助变量。
    total_costs = 0.0
    iters = 0
    state = session.run(model.initial_state)
    # 训练一个epoch。
    for step,(x,y) in enumerate(reader.ptb_iterator(data,model.batch_size,model.num_steps)):
        #在当前batch上运行train_op并计算损失值。交叉熵损失函数计算的就是下一个单词为给定单词的概率
        cost,state,_ = session.run([model.cost,model.final_state,train_op],
                                   {model.input_data:x,model.targets:y,
                                    model.initial_state:state})
        #将不同时刻，不同batch的概率加起来就可以得到第二个perplexity公司等号右边的部分，
        #再将这个和做指数运算就可以得到perplexity值
        total_costs+=cost
        iters+=model.num_steps
        
        #只有在训练时输出日志
        if output_log and step % 100 == 0:
            print("after % step ,perplexity is %.3f" %(step,np.exp(total_costs/iters)))
        
        #返回给定模型在给定数据上的perplexity值
        return np.exp(total_costs/iters)
    
def main(_):
    # 获取原始数据
    train_data, valid_data, test_data, _ = reader.ptb_raw_data(DATA_PATH)
    
    # 计算一个epoch需要训练的次数
    #train_data_len = len(train_data)
    #train_batch_len = train_data_len  # # TRAIN_BATCH_SIZE
    #train_epoch_size = (train_batch_len - 1)  # # TRAIN_NUM_STEP

    #valid_data_len = len(valid_data)
    #valid_batch_len = valid_data_len  # # EVAL_BATCH_SIZE
    #valid_epoch_size = (valid_batch_len - 1)  # # EVAL_NUM_STEP

    #test_data_len = len(test_data)
    #test_batch_len = test_data_len  # # EVAL_BATCH_SIZE
    #test_epoch_size = (test_batch_len - 1)  # # EVAL_NUM_STEP
    
    # 定义初始化函数
    initializer = tf.random_uniform_initializer(-0.05, 0.05)
    
    with tf.variable_scope("language_model", reuse=None, initializer=initializer):
        train_model = PTBModel(True, TRAIN_BATCH_SIZE, TRAIN_NUM_STEP)
    
    # 定义评测用的循环神经网络模型
    with tf.variable_scope("language_model", reuse=True, initializer=initializer):
        eval_model = PTBModel(False, EVAL_BATCH_SIZE, EVAL_NUM_STEP)
        
    with tf.Session() as session:
        tf.global_variables_initializer().run()
        
        # 使用训练数据训练模型
        for i in range(NUM_EPOCH):
            print("In iteration:%d" % (i + 1))
            # 在所有训练数据上训练循环神经网络模型
            run_epoch(session, train_model, train_data, train_model.train_op, True)
            
            # 使用验证数据评测模型效果
            valid_perplexity = run_epoch(session, eval_model, valid_data, tf.no_op(), False)
            print("Epoch: %d Validation Perplexity : %.3f" % (i + 1, valid_perplexity))
            
        # 最后使用测试数据测试模型效果
        test_perplexity = run_epoch(session, eval_model, test_data, tf.no_op(), False)
        print("Test Perplexity:%.3f" % test_perplexity)
    
if __name__ == "__main__":
    tf.app.run()
        

# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================


"""Utilities for parsing PTB text files."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import collections
import os

import numpy as np
import tensorflow as tf


def _read_words(filename):
  with tf.gfile.GFile(filename, "r") as f:
    return f.read().replace("\n", "<eos>").split() #读取文件， 将换行符替换为 <eos>， 然后将文件按空格分割。 返回一个 1-D list


def _build_vocab(filename):  #用于建立字典
  data =http://www.mamicode.com/ _read_words(filename)
  counter = collections.Counter(data) #输出一个字典： key是word， value是这个word出现的次数
  count_pairs = sorted(counter.items(), key=lambda x: (-x[1], x[0]))
#counter.items() 会返回一个tuple列表， tuple是(key, value), 按 value的降序，key的升序排列
  words, _ = list(zip(*count_pairs)) #感觉这个像unzip 就是把key放在一个tuple里，value放在一个tuple里
  word_to_id = dict(zip(words, range(len(words))))#对每个word进行编号， 按照之前words输出的顺序（value降序，key升序）
  return word_to_id  #返回dict， key：word， value：id


def _file_to_word_ids(filename, word_to_id): #将file表示为word_id的形式
  data =http://www.mamicode.com/ _read_words(filename)
  return [word_to_id[word] for word in data]

def ptb_raw_data(data_path=None):
  """Load PTB raw data from data directory "data_path".
  Reads PTB text files, converts strings to integer ids,
  and performs mini-batching of the inputs.
  The PTB dataset comes from Tomas Mikolov‘s webpage:
  http://www.fit.vutbr.cz/~imikolov/rnnlm/simple-examples.tgz
  Args:
    data_path: string path to the directory where simple-examples.tgz has
      been extracted.
  Returns:
    tuple (train_data, valid_data, test_data, vocabulary)
    where each of the data objects can be passed to PTBIterator.
  """

  train_path = os.path.join(data_path, "ptb.train.txt")
  valid_path = os.path.join(data_path, "ptb.valid.txt")
  test_path = os.path.join(data_path, "ptb.test.txt")

  word_to_id = _build_vocab(train_path) #使用训练集确定word id
  train_data =http://www.mamicode.com/ _file_to_word_ids(train_path, word_to_id)
  valid_data = _file_to_word_ids(valid_path, word_to_id)
  test_data = _file_to_word_ids(test_path, word_to_id)
  vocabulary = len(word_to_id)#字典的大小
  return train_data, valid_data, test_data, vocabulary


def ptb_iterator(raw_data, batch_size, num_steps):
  """Iterate on the raw PTB data.
  This generates batch_size pointers into the raw PTB data, and allows
  minibatch iteration along these pointers.
  Args:
    raw_data: one of the raw data outputs from ptb_raw_data.
    batch_size: int, the batch size.
    num_steps: int, the number of unrolls.
  Yields:
    Pairs of the batched data, each a matrix of shape [batch_size, num_steps].
    The second element of the tuple is the same data time-shifted to the
    right by one.
  Raises:
    ValueError: if batch_size or num_steps are too high.
  """
  raw_data = np.array(raw_data, dtype=np.int32)#raw data : train_data | vali_data | test data

  data_len = len(raw_data) #how many words in the data_set
  batch_len = data_len // batch_size
  data = np.zeros([batch_size, batch_len], dtype=np.int32)#batch_len 就是几个word的意思
  for i in range(batch_size):
    data[i] = raw_data[batch_len * i:batch_len * (i + 1)]

  epoch_size = (batch_len - 1) // num_steps

  if epoch_size == 0:
    raise ValueError("epoch_size == 0, decrease batch_size or num_steps")

  for i in range(epoch_size):
    x = data[:, i*num_steps:(i+1)*num_steps]
    y = data[:, i*num_steps+1:(i+1)*num_steps+1]
  yield (x, y)