介紹

上一篇我們講了如何在PyTorch框架下用RNN分類名字http://blog.csdn.net/m0_37306360/article/details/79316013，本文講如何用RNN生成特定語言(類別)的名字。我們使用上一篇同樣的資料集。不同的是，不是根據輸入的名字來預測此名字是那種語言的(讀完名字的所有字母之後，我們不是預測一個類別)。而是一次輸入一個類別並輸出一個字母。 反覆預測字元以生成名字。

PyTorch之RNN實戰生成

定義網路

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size)
 
:
        super(RNN, self).__init__()
        self.hidden_size = hidden_size

        self.i2h = nn.Linear(n_categories + input_size + hidden_size, hidden_size)
        self.i2o = nn.Linear(n_categories + input_size + hidden_size, output_size)
        self.o2o = nn.Linear(hidden_size + output_size, output_size)
        self.dropout = nn.Dropout(0.1
 
)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, category, input, hidden):
        input_combined = torch.cat((category, input, hidden), 1)
        output = self.i2o(input_combined)
        hidden = self.i2h(input_combined)
        output_combined = torch.cat((output, hidden), 1)
        output = self.o2o(output_combined)
        output = self.dropout(output)
        output = self.softmax(output)
        return
 
 output, hidden

    def initHidden(self):
        return Variable(torch.zeros(1, self.hidden_size))

資料預處理

其他的資料預處理和上一篇文章類似。主要的不同是如何構建訓練樣本對。

對於每個時間步（即訓練詞中的每個字母），網路的輸入是（類別，當前字母，隱藏狀態），輸出將是（下一個字母，下一個隱藏狀態）。 因此，對於每個訓練集，我們需要：類別，一組輸入字母和一組輸出(目標)字母。

我們可以很簡單的獲取(類別，名字(字母序列))，但是我們如何用這些資料構建輸入字母序列和輸出字母序列呢？由於我們預測了每個時間步的當前字母的下一個字母，因此字母對是來自行的連續字母的組 - 例如對於序列(“ABCD”)，構建(“A”, B”), (“B”,“C”), (“C”, “D”), (“D”,“EOS”). 如圖：

訓練網路

與僅使用最後一個輸出的分類相比，我們在每一步都進行了預測，因此我們需要計算每一步的損失。

for i in range(input_line_tensor.size()[0]):
        output, hidden = rnn(category_tensor, input_line_tensor[i], hidden)
        loss += criterion(output, target_line_tensor[i])

取樣

為了取樣，我們給網路輸入一個字母，得到下一個字母，把它作為下一個字母的輸入，並重復，直到返回EOS結束符。

整個生成過程如下：
1. 為輸入類別，開始字母和空的隱藏狀態建立張量
2. 用開始字母建立一個字串:output_name
3. 達到最大輸出長度:
(1).將當前的字母送入網路
(2).從輸出獲取下一個字母和下一個隱藏狀態
(3).如果這個字元是EOS，生成結束
(4).如果是普通字母，新增到output_name並繼續
4.返回最後生成的名字

完整程式碼

from io import open
import glob
import unicodedata
import string

all_letters = string.ascii_letters + " .,;'-"
n_letters = len(all_letters) + 1 # Plus EOS marker

def findFiles(path): return glob.glob(path)

# Turn a Unicode string to plain ASCII, thanks to http://stackoverflow.com/a/518232/2809427
def unicodeToAscii(s):
    return ''.join(
        c for c in unicodedata.normalize('NFD', s)
        if unicodedata.category(c) != 'Mn'
        and c in all_letters
    )

# Read a file and split into lines
def readLines(filename):
    lines = open(filename, encoding='utf-8').read().strip().split('\n')
    return [unicodeToAscii(line) for line in lines]

# Build the category_lines dictionary, a list of lines per category
category_lines = {}
all_categories = []
for filename in findFiles('nlpdata/data/names/*.txt'):
    category = filename.split('/')[-1].split('.')[0]
    category = category.split('\\')[1]
    all_categories.append(category)
    lines = readLines(filename)
    category_lines[category] = lines

n_categories = len(all_categories)

# print('# categories:', n_categories, all_categories)
# print(unicodeToAscii("O'Néàl"))

import torch
import torch.nn as nn
from torch.autograd import Variable

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()
        self.hidden_size = hidden_size
        self.i2h = nn.Linear(n_categories + input_size + hidden_size, hidden_size)
        self.i2o = nn.Linear(n_categories + input_size + hidden_size, output_size)
        self.o2o = nn.Linear(hidden_size + output_size, output_size)
        self.dropout = nn.Dropout(0.1)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, category, input, hidden):
        input_combined = torch.cat((category, input, hidden), 1)
        output = self.i2o(input_combined)
        hidden = self.i2h(input_combined)
        output_combined = torch.cat((output, hidden), 1)
        output = self.o2o(output_combined)
        output = self.dropout(output)
        output = self.softmax(output)
        return output, hidden

    def initHidden(self):
        return Variable(torch.zeros(1, self.hidden_size))

# Training

# Preparing for Training (get random pairs of (category, line))
import random

# Random item from a list
def randomChoice(l):
    return l[random.randint(0, len(l) - 1)]

# Get a random category and random line from that category
def randomTrainingPair():
    category = randomChoice(all_categories)
    line = randomChoice(category_lines[category])
    return category, line

# One-hot vector for category
def categoryTensor(category):
    li = all_categories.index(category)
    tensor = torch.zeros(1, n_categories) # 1*18
    tensor[0][li] = 1
    return tensor

# One-hot matrix of first to last letters (not including EOS) for input
def inputTensor(line):
    tensor = torch.zeros(len(line), 1, n_letters) # len(line)*1*59
    for li in range(len(line)):
        letter = line[li]
        tensor[li][0][all_letters.find(letter)] = 1
    return tensor

# LongTensor of second letter to end (EOS) for target
def targetTensor(line):
    letter_indexes = [all_letters.find(line[li]) for li in range(1, len(line))]
    letter_indexes.append(n_letters - 1) # EOS
    return torch.LongTensor(letter_indexes)

# Make category, input, and target tensors from a random category, line pair
def randomTrainingExample():
    category, line = randomTrainingPair()
    category_tensor = Variable(categoryTensor(category))
    input_line_tensor = Variable(inputTensor(line))
    target_line_tensor = Variable(targetTensor(line))
    return category_tensor, input_line_tensor, target_line_tensor

# our model
rnn = RNN(n_letters, 128, n_letters)

# Training the Network

criterion = nn.NLLLoss()
learning_rate = 0.0005

def train(category_tensor, input_line_tensor, target_line_tensor):
    hidden = rnn.initHidden()

    rnn.zero_grad()
    loss = 0

    for i in range(input_line_tensor.size()[0]):
        output, hidden = rnn(category_tensor, input_line_tensor[i], hidden)
        loss += criterion(output, target_line_tensor[i])

    loss.backward()

    for p in rnn.parameters():
        p.data.add_(-learning_rate, p.grad.data)

    return output, loss.data[0] / input_line_tensor.size()[0]


import time
import math

def timeSince(since):
    now = time.time()
    s = now - since
    m = math.floor(s / 60)
    s -= m * 60
    return '%dm %ds' % (m, s)

n_iters = 100000
print_every = 5000
plot_every = 500
all_losses = []
total_loss = 0 # Reset every plot_every iters

start = time.time()


for iter in range(1, n_iters + 1):
    output, loss = train(*randomTrainingExample())
    total_loss += loss

    if iter % print_every == 0:
        print('%s (%d %d%%) %.4f' % (timeSince(start), iter, iter / n_iters * 100, loss))

    if iter % plot_every == 0:
        all_losses.append(total_loss / plot_every)
        total_loss = 0

# 訓練loss變化
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker

plt.figure()
plt.plot(all_losses)
plt.show()

# Sample network
max_length = 20
# Sample from a category and starting letter
def sample(category, start_letter='A'):
    category_tensor = Variable(categoryTensor(category))
    input = Variable(inputTensor(start_letter))
    hidden = rnn.initHidden()

    output_name = start_letter

    # 最大也只生成max_length長度
    for i in range(max_length):
        output, hidden = rnn(category_tensor, input[0], hidden)
        topv, topi = output.data.topk(1)
        topi = topi[0][0]
        # 如果是EOS，停止
        if topi == n_letters - 1:
            break
        else:
            letter = all_letters[topi]
            output_name += letter
        # 否則，將這個時刻輸出的字母作為下個時刻的輸入字母
        input = Variable(inputTensor(letter))

    return output_name

# Get multiple samples from one category and multiple starting letters
def samples(category, start_letters='ABC'):
    for start_letter in start_letters:
        print(sample(category, start_letter))

samples('Russian', 'RUS')
samples('German', 'GER')
samples('Spanish', 'SPA')
samples('Chinese', 'CHI')

輸出結果：
0m 32s (5000 5%) 2.3348
1m 1s (10000 10%) 3.0012
1m 29s (15000 15%) 2.7776
2m 2s (20000 20%) 2.7482
2m 38s (25000 25%) 1.3141
3m 10s (30000 30%) 2.5318
3m 37s (35000 35%) 2.4345
4m 3s (40000 40%) 2.7806
4m 30s (45000 45%) 2.0744
4m 57s (50000 50%) 2.7273
5m 23s (55000 55%) 5.1529
5m 49s (60000 60%) 2.0862
6m 15s (65000 65%) 2.5506
6m 41s (70000 70%) 3.4072
7m 8s (75000 75%) 2.6554
7m 34s (80000 80%) 2.1122
8m 12s (85000 85%) 2.2132
8m 44s (90000 90%) 1.9226
9m 11s (95000 95%) 2.8443
9m 37s (100000 100%) 2.4129
Roskin
Uakinov
Santovov
Gerran
Eren
Romer
Sara
Parez
Aller
Chan
Han
Iun

Process finished with exit code 0

loss變化：