【PyTorch】循环神经网络

循环神经网络是什么

Recurrent Neural Networks

RNN:循环神经网络

  • 处理不定长输入的模型
  • 常用于NLP及时间序列任务(输入数据具有前后关系)

RNN网络结构

参考资料
Recurrent Neural Networks Tutorial, Part 1 -- Introduction to RNNs
Understanding LSTM Networks

RNN实现人名分类

问题定义:输入任意长度姓名(字符串),输出姓名来自哪一个国家(18类分类任务)

数据: https://download.pytorch.org/tutorial/data.zip

Jackie Chan ------ 成龙

Jay Chou ------ 周杰伦

Tingsong Yue ------ 余霆嵩

RNN如何处理不定长输入

思考:计算机如何实现不定长字符串 到分类向量 的映射?

Chou(字符串)→ RNN →Chinese(分类类别)

  1. 单词字符 → 数字

  2. 数字 → model

  3. 下一个字符 → 数字 → model

  4. 最后一个字符 → 数字 → model → 分类向量

    伪代码

    Chou(字符串)→ RNN →Chinese(分类类别)

    for string in [C, h, o, u]:
    1. one-hot:string → [0,0, ...., 1, ..., 0] # 首先把每个字母转换成编码
    2. y, h = model([0,0, ...., 1, ..., 0], h) # h就是隐藏层的状态信息

xt:时刻t的输入,shape = (1, 57)

st:时刻t的状态值,shape=(1, 128)

ot:时刻t的输出值,shape=(1, 18)

U:linear层的权重参数, shape = (128, 57)

W:linear层的权重参数, shape = (128, 128)

V:linear层的权重参数, shape = (18, 128)

代码如下:

python 复制代码
# -*- coding: utf-8 -*-
"""
# @file name  : rnn_demo.py
# @author     : TingsongYu https://github.com/TingsongYu
# @date       : 2019-12-09
# @brief      : rnn人名分类
"""
from io import open
import glob
import unicodedata
import string
import math
import os
import time
import torch.nn as nn
import torch
import random
import matplotlib.pyplot as plt
import torch.utils.data
import sys
# 获取路径
hello_pytorch_DIR = os.path.abspath(os.path.dirname(__file__)+os.path.sep+".."+os.path.sep+"..")
sys.path.append(hello_pytorch_DIR)

from tools.common_tools import set_seed

set_seed(1)  # 设置随机种子
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
# 选择运行设备
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device = torch.device("cpu")


# 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]


def unicodeToAscii(s):
    return ''.join(
        c for c in unicodedata.normalize('NFD', s)
        if unicodedata.category(c) != 'Mn'
        and c in all_letters)


# Find letter index from all_letters, e.g. "a" = 0
def letterToIndex(letter):
    return all_letters.find(letter)


# Just for demonstration, turn a letter into a <1 x n_letters> Tensor
def letterToTensor(letter):
    tensor = torch.zeros(1, n_letters)
    tensor[0][letterToIndex(letter)] = 1
    return tensor


# Turn a line into a <line_length x 1 x n_letters>,
# or an array of one-hot letter vectors
def lineToTensor(line):
    tensor = torch.zeros(len(line), 1, n_letters)
    for li, letter in enumerate(line):
        tensor[li][0][letterToIndex(letter)] = 1
    return tensor


def categoryFromOutput(output):
    top_n, top_i = output.topk(1)
    category_i = top_i[0].item()
    return all_categories[category_i], category_i


def randomChoice(l):
    return l[random.randint(0, len(l) - 1)]


def randomTrainingExample():
    category = randomChoice(all_categories)                 # 选类别
    line = randomChoice(category_lines[category])           # 选一个样本
    category_tensor = torch.tensor([all_categories.index(category)], dtype=torch.long)
    line_tensor = lineToTensor(line)    # str to one-hot
    return category, line, category_tensor, line_tensor


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


# Just return an output given a line
def evaluate(line_tensor):
    hidden = rnn.initHidden()

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    return output


def predict(input_line, n_predictions=3):
    print('\n> %s' % input_line)
    with torch.no_grad():
        output = evaluate(lineToTensor(input_line))

        # Get top N categories
        topv, topi = output.topk(n_predictions, 1, True)

        for i in range(n_predictions):
            value = topv[0][i].item()
            category_index = topi[0][i].item()
            print('(%.2f) %s' % (value, all_categories[category_index]))


def get_lr(iter, learning_rate):
    lr_iter = learning_rate if iter < n_iters else learning_rate*0.1
    return lr_iter

# 定义网络结构
class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()

        self.hidden_size = hidden_size

        self.u = nn.Linear(input_size, hidden_size)
        self.w = nn.Linear(hidden_size, hidden_size)
        self.v = nn.Linear(hidden_size, output_size)

        self.tanh = nn.Tanh()
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, inputs, hidden):

        u_x = self.u(inputs)

        hidden = self.w(hidden)
        hidden = self.tanh(hidden + u_x)

        output = self.softmax(self.v(hidden))

        return output, hidden

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


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

    rnn.zero_grad()

    line_tensor = line_tensor.to(device)
    hidden = hidden.to(device)
    category_tensor = category_tensor.to(device)

    for i in range(line_tensor.size()[0]):
        output, hidden = rnn(line_tensor[i], hidden)

    loss = criterion(output, category_tensor)
    loss.backward()

    # Add parameters' gradients to their values, multiplied by learning rate
    for p in rnn.parameters():
        # p.data.add_(-learning_rate, p.grad.data) # 该方法已经被弃用
        p.data.add_(p.grad.data, alpha=-learning_rate)

    return output, loss.item()


if __name__ == "__main__":
    print(device)

    # config
    data_dir = os.path.abspath(os.path.join(BASE_DIR, "..", "..", "data", "rnn_data", "names"))
    if not os.path.exists(data_dir):
        raise Exception("\n{} 不存在,请下载 08-05-数据-20200724.zip  放到\n{}  下,并解压即可".format(
            data_dir, os.path.dirname(data_dir)))

    path_txt = os.path.join(data_dir, "*.txt")
    all_letters = string.ascii_letters + " .,;'"
    n_letters = len(all_letters)    # 52 + 5 字符总数
    print_every = 5000
    plot_every = 5000
    learning_rate = 0.005
    n_iters = 200000

    # step 1 data
    # Build the category_lines dictionary, a list of names per language
    category_lines = {}
    all_categories = []
    for filename in glob.glob(path_txt):
        category = os.path.splitext(os.path.basename(filename))[0]
        all_categories.append(category)
        lines = readLines(filename)
        category_lines[category] = lines

    n_categories = len(all_categories)

    # step 2 model
    n_hidden = 128
    # rnn = RNN(n_letters, n_hidden, n_categories)
    rnn = RNN(n_letters, n_hidden, n_categories)

    rnn.to(device)

    # step 3 loss
    criterion = nn.NLLLoss()

    # step 4 optimize by hand

    # step 5 iteration
    current_loss = 0
    all_losses = []
    start = time.time()
    for iter in range(1, n_iters + 1):
        # sample
        category, line, category_tensor, line_tensor = randomTrainingExample()

        # training
        output, loss = train(category_tensor, line_tensor)

        current_loss += loss

        # Print iter number, loss, name and guess
        if iter % print_every == 0:
            guess, guess_i = categoryFromOutput(output)
            correct = '✓' if guess == category else '✗ (%s)' % category
            print('Iter: {:<7} time: {:>8s} loss: {:.4f} name: {:>10s}  pred: {:>8s} label: {:>8s}'.format(
                iter, timeSince(start), loss, line, guess, correct))

        # Add current loss avg to list of losses
        if iter % plot_every == 0:
            all_losses.append(current_loss / plot_every)
            current_loss = 0

path_model = os.path.abspath(os.path.join(BASE_DIR, "..", "..", "data", "rnn_state_dict.pkl"))
if not os.path.exists(path_model):
    raise Exception("\n{} 不存在,请下载 08-05-数据-20200724.zip  放到\n{}  下,并解压即可".format(
        path_model, os.path.dirname(path_model)))
torch.save(rnn.state_dict(), path_model)
plt.plot(all_losses)
plt.show()

predict('Yue Tingsong')
predict('Yue tingsong')
predict('yutingsong')

predict('test your name')
相关推荐
吴佳浩5 小时前
我把 HAMi 摆回一台双卡机器上跑了一遍:GPU 编排的性能账,到底记在谁头上
人工智能·gpu
forest_099 小时前
《AI 编程 Harness 实战手册》04|审批这道门:点了一百次“允许”,还剩多少作用?
人工智能·算法·prompt·agi
火山引擎开发者社区10 小时前
当 RPA 遇上 Agent:MUA 混合执行 Workflow 深度拆解
人工智能
严肃小白兔10 小时前
第 10/24 期:深度神经网络的自动求导机制与梯度验证
人工智能·神经网络·dnn
火山引擎开发者社区10 小时前
定档12.16-12.18!FORCE原动力大会上海见!
人工智能
努力的MM12311 小时前
CAIE 二级和阿里云大模型 ACP 怎么选?企业 AI 工程与云平台开发能力对比
人工智能
老金带你玩AI11 小时前
Kling Image 4.0来了!8K和组图直出,电商人的提效神器?
人工智能
孟健11 小时前
用 unslop 拯救 AI 生成网页的塑料感,7 分钟重构落地实测
前端·人工智能·ai编程
stormzhangV11 小时前
Google 重返第一梯队
人工智能·google·aigc