本文将演示如何通过构建 Gated Recurrent Unit Network 来构建文本生成器。训练网络的概念过程是首先向网络提供网络正在训练的文本中存在的每个字符的映射到唯一数字。然后将每个字符热编码为一个向量，这是网络所需的格式。
所述程序的数据是著名诗人的短诗和著名诗集，格式为 .txt。它可以从 kaggle 下载。
第 1 步：导入所需的库
- Python3 语言
-
本文将演示如何通过构建 Gated Recurrent Unit Network 来构建文本生成器。训练网络的概念过程是首先向网络提供网络正在训练的文本中存在的每个字符的映射到唯一数字。然后将每个字符热编码为一个向量，这是网络所需的格式。 所述程序的数据是著名诗人的短诗和著名诗集，格式为 .txt。它可以从 kaggle 下载。 第 1 步：导入所需的库 from __future__ import absolute_import, division, print_function, unicode_literals import numpy as np import tensorflow as tf from keras.models import Sequential from keras.layers import Dense, Activation from keras.layers import LSTM from keras.optimizers import RMSprop from keras.callbacks import LambdaCallback from keras.callbacks import ModelCheckpoint from keras.callbacks import ReduceLROnPlateau import random import sys 第 2 步：将数据加载到字符串中 # Changing the working location to the location of the text file cd C:\Users\Dev\Desktop\Kaggle\Poems # Reading the text file into a string with open('poems.txt', 'r') as file: text = file.read() # A preview of the text file print(text) 第 3 步：创建从文本中的每个唯一字符到唯一数字的映射 # Storing all the unique characters present in the text vocabulary = sorted(list(set(text))) # Creating dictionaries to map each character to an index char_to_indices = dict((c, i) for i, c in enumerate(vocabulary)) indices_to_char = dict((i, c) for i, c in enumerate(vocabulary)) print(vocabulary) 步骤 4：预处理数据 # Dividing the text into subsequences of length max_length # So that at each time step the next max_length characters # are fed into the network max_length = 100 steps = 5 sentences = [] next_chars = [] for i in range(0, len(text) - max_length, steps): sentences.append(text[i: i + max_length]) next_chars.append(text[i + max_length]) # Hot encoding each character into a boolean vector # Initializing a matrix of boolean vectors with each column representing # the hot encoded representation of the character X = np.zeros((len(sentences), max_length, len(vocabulary)), dtype = np.bool) y = np.zeros((len(sentences), len(vocabulary)), dtype = np.bool) # Placing the value 1 at the appropriate position for each vector # to complete the hot-encoding process for i, sentence in enumerate(sentences): for t, char in enumerate(sentence): X[i, t, char_to_indices[char]] = 1 y[i, char_to_indices[next_chars[i]]] = 1 第 5 步：构建 GRU 网络 # Initializing the LSTM network model = Sequential() # Defining the cell type model.add(GRU(128, input_shape =(max_length, len(vocabulary)))) # Defining the densely connected Neural Network layer model.add(Dense(len(vocabulary))) # Defining the activation function for the cell model.add(Activation('softmax')) # Defining the optimizing function optimizer = RMSprop(lr = 0.01) # Configuring the model for training model.compile(loss ='categorical_crossentropy', optimizer = optimizer) 第 6 步：定义一些将在网络训练期间使用的辅助函数请注意，下面给出的前两个函数引用自 Keras 团队的官方文本生成示例的文档。 a） 用于对下一个字符进行采样的辅助函数： # Helper function to sample an index from a probability array def sample_index(preds, temperature = 1.0): # temperature determines the freedom the function has when generating text # Converting the predictions vector into a numpy array preds = np.asarray(preds).astype('float64') # Normalizing the predictions array preds = np.log(preds) / temperature exp_preds = np.exp(preds) preds = exp_preds / np.sum(exp_preds) # The main sampling step. Creates an array of probabilities signifying # the probability of each character to be the next character in the # generated text probas = np.random.multinomial(1, preds, 1) # Returning the character with maximum probability to be the next character # in the generated text return np.argmax(probas) b） 辅助函数，用于在每个 epoch 后生成文本 # Helper function to generate text after the end of each epoch def on_epoch_end(epoch, logs): print() print('----- Generating text after Epoch: % d' % epoch) # Choosing a random starting index for the text generation start_index = random.randint(0, len(text) - max_length - 1) # Sampling for different values of diversity for diversity in [0.2, 0.5, 1.0, 1.2]: print('----- diversity:', diversity) generated = '' # Seed sentence sentence = text[start_index: start_index + max_length] generated += sentence print('----- Generating with seed: "' + sentence + '"') sys.stdout.write(generated) for i in range(400): # Initializing the predictions vector x_pred = np.zeros((1, max_length, len(vocabulary))) for t, char in enumerate(sentence): x_pred[0, t, char_to_indices[char]] = 1. # Making the predictions for the next character preds = model.predict(x_pred, verbose = 0)[0] # Getting the index of the most probable next character next_index = sample_index(preds, diversity) # Getting the most probable next character using the mapping built next_char = indices_to_char[next_index] # Building the generated text generated += next_char sentence = s