From 0325398a8b994f3d72fc387b775510c04fe53e82 Mon Sep 17 00:00:00 2001 From: SivanLaai Date: Mon, 18 Dec 2023 17:45:09 +0800 Subject: [PATCH] update passages --- content/posts/learning/rl/actor_critic.md | 34 +- content/posts/learning/rl/policy_gradient.md | 3 +- .../posts/technology/baidu_dict_crawler.md | 466 ++++++++++++++++++ 3 files changed, 492 insertions(+), 11 deletions(-) create mode 100644 content/posts/technology/baidu_dict_crawler.md diff --git a/content/posts/learning/rl/actor_critic.md b/content/posts/learning/rl/actor_critic.md index 9930345..4651179 100644 --- a/content/posts/learning/rl/actor_critic.md +++ b/content/posts/learning/rl/actor_critic.md @@ -1,5 +1,5 @@ --- -title: 演员评论家算法 +title: 强化学习A2C算法解析与实现 date: 2023-10-07T14:57:55+08:00 draft: true categories: @@ -27,6 +27,7 @@ tags: 这样做的好处就是，大于b的Reward对应的动作概率就会增加，而小于b的Reward对应的动作就会减小，这就可以在一定程度上避免没有采样的概率变小。 +# 概念解释 ## 批评家Critic 批评家不作什么动作，而是对演员的动作进行打分。 @@ -39,8 +40,15 @@ tags: 学习一个策略函数，根据状态s选择下一步的动作。 -# 算法 +## 算法流程 +# 算法Python实现和解释 + +### 前置需求 +- 安装numpy +- 安装torch cuda版本 +- 强化学习gym环境 +### 导入必要的包 ```python import numpy as np @@ -52,18 +60,21 @@ import torch.nn.functional as F import torch from torch import optim from torch import Tensor - +# 定义需要的包 device = torch.device("cuda" if torch.cuda.is_available else "cpu") - +``` +### 实现对数熵 +```python class SoftCrossEntryopy(nn.CrossEntropyLoss): def __init__(self, weight: Tensor = None, size_average=None, ignore_index: int = -100, reduce=None, reduction: str = 'mean', label_smoothing: float = 0.0) -> None: super().__init__(weight, size_average, reduce, reduction) def forward(self, input: Tensor, target: Tensor) -> Tensor: - #output = torch.mean(torch.sum(-torch.log(input)*target), 1) output = torch.mean(torch.sum(-torch.log(input)*target)) return output +``` +```python class Critic(nn.Module): def __init__(self, n_features, @@ -86,7 +97,9 @@ class Critic(nn.Module): #x = self.ln(x) x = self.fc2(x) return x +``` +```python class Actor(nn.Module): def __init__(self, n_actions, @@ -111,7 +124,9 @@ class Actor(nn.Module): x = self.fc2(x) x = F.softmax(x) # action probs return x +``` +```python class AdvancedActorCritic: def __init__( self, @@ -195,15 +210,17 @@ class AdvancedActorCritic: self.actor_learn(s, a, td_error) +``` +```python import gym import matplotlib.pyplot as plt -DISPLAY_REWARD_THRESHOLD = 50 # renders environment if total episode reward is greater then this threshold -RENDER = False # rendering wastes time +DISPLAY_REWARD_THRESHOLD = 50 +RENDER = False env = gym.make('CartPole-v0') -env.seed(1) # reproducible, general Policy gradient has high variance +env.seed(1) env = env.unwrapped print(env.action_space) @@ -216,7 +233,6 @@ RL = AdvancedActorCritic( n_features=env.observation_space.shape[0], learning_rate=0.02, reward_decay=0.99, - # output_graph=True, ) for i_episode in range(3000): diff --git a/content/posts/learning/rl/policy_gradient.md b/content/posts/learning/rl/policy_gradient.md index 1ffa7af..8541525 100644 --- a/content/posts/learning/rl/policy_gradient.md +++ b/content/posts/learning/rl/policy_gradient.md @@ -1,5 +1,5 @@ --- -title: 强化学习策略梯度 +title: 强化学习策略梯度算法解析与代码实现 date: 2023-11-22T16:14:17+08:00 draft: true categories: @@ -148,7 +148,6 @@ RL = PolicyGradient( n_features=env.observation_space.shape[0], learning_rate=0.02, reward_decay=0.99, - # output_graph=True, ) for i_episode in range(3000): diff --git a/content/posts/technology/baidu_dict_crawler.md b/content/posts/technology/baidu_dict_crawler.md new file mode 100644 index 0000000..31b0c08 --- /dev/null +++ b/content/posts/technology/baidu_dict_crawler.md @@ -0,0 +1,466 @@ +--- +title: 百度汉语词典爬虫 +date: 2023-12-18T16:32:53+08:00 +draft: true +categories: +tags: +--- +# 准备工作 + +### 分析请求头信息 +- 为了防止限制，打开[百度汉语官网](https://hanyu.baidu.com/)并登录，使用浏览器F12复制所有请求头信息，其中cookie是登录信息 +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218171154.png) + +### 分析根据关键字查找所有词语的列表接口 +- 在百度汉语主页面搜索关键字“诗词大全”，会以列表的形式展现所有相关的词语， +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218172801.png) +- 使用F12或者是其他的抓包工具，分析到新[接口](https://hanyu.baidu.com/hanyu/ajax/search_list)会返回对应的词典信息，如下： +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218172530.png) + +- 分析参数 +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218172727.png) +wd 是 包含的关键字的所有词语，pn表示页码，其他参数不是很重要，实际测试发现，不设置也返回，这里建议保留你浏览器生成的默认参数。 +### 在mysql中建如下表格 + +```mysql +CREATE TABLE `single_character_info` ( + `pinyin` varchar(255) CHARACTER SET utf8mb4 NOT NULL, + `word` varchar(255) NOT NULL, + `plainPinyin` varchar(255) CHARACTER SET utf8mb4 NOT NULL, + `definition` varchar(4096) CHARACTER SET utf8mb4 DEFAULT NULL, + `pronunciation` varchar(255) DEFAULT NULL, + `wordID` int DEFAULT NULL, + PRIMARY KEY (`word`,`pinyin`) USING BTREE +) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4; +``` + +### 准备单个词典文件 + +下载地址：[BaiduPinyinCrawler/data/clover.base.dict.yaml at main · SivanLaai/BaiduPinyinCrawler (github.com)](https://github.com/SivanLaai/BaiduPinyinCrawler/blob/main/data/clover.base.dict.yaml) + +### 安装python对应安装包 + +- 使用pip安装如下包： + +```bash +certifi==2022.6.15 +charset-normalizer==2.1.1 +idna==3.3 +PyMySQL==1.0.2 +requests==2.28.1 +six==1.16.0 +urllib3==1.26.12 +``` + +# 代码实现 + +### 日志模块 +```python +# Logger.py +import logging, logging.handlers, time, os +from Config import config + + +class Log(object): + """ + logging的初始化操作，以类封装的形式进行 + """ + + # 设置输出的等级 + LEVELS = { + "NOSET": logging.NOTSET, + "DEBUG": logging.DEBUG, + "INFO": logging.INFO, + "WARNING": logging.WARNING, + "ERROR": logging.ERROR, + "CRITICAL": logging.CRITICAL, + } + log_level = config['LOG']['LEVEL'] + + def __init__(self): + # 定义对应的程序模块名name，默认为root + self.logger = logging.getLogger() + # 每次被调用后，清空已经存在handler + self.logger.handlers.clear() + + # log_path是存放日志的路径 + service_name = "FundCrawler" + timestr = time.strftime("%Y_%m_%d", time.localtime(time.time())) + lib_path = config['LOG']['LOG_PATH'] + # 如果不存在这个logs文件夹，就自动创建一个 + if not os.path.exists(lib_path): + os.mkdir(lib_path) + # 日志文件的地址 + self.logname = lib_path + "/" + service_name + "_" + timestr + ".log" + + # 必须设置，这里如果不显示设置，默认过滤掉warning之前的所有级别的信息 + self.logger.setLevel(self.log_level) + + # 日志输出格式 + self.formatter = logging.Formatter( + "[%(asctime)s] [%(levelname)s]: %(message)s" + ) # [2019-05-15 14:48:52,947] - test.py] - ERROR: this is error + + # 创建一个FileHandler， 向文件logname输出日志信息 + # fh = logging.FileHandler(self.logname, 'a', encoding='utf-8') + fh = logging.handlers.RotatingFileHandler( + filename=self.logname, maxBytes=1024 * 1024 * 50, backupCount=5 + ) + # 设置日志等级 + fh.setLevel(self.log_level) + # 设置handler的格式对象 + fh.setFormatter(self.formatter) + # 将handler增加到logger中 + self.logger.addHandler(fh) + + # 创建一个StreamHandler,用于输出到控制台 + ch = logging.StreamHandler() + ch.setLevel(self.log_level) + ch.setFormatter(self.formatter) + self.logger.addHandler(ch) + + # # 关闭打开的文件 + fh.close() + + def info(self, message): + self.logger.info(message) + + def debug(self, message): + self.logger.debug(message) + + def warning(self, message): + self.logger.warning(message) + + def error(self, message): + self.logger.error(message) + + def critical(self, message): + self.logger.critical(message) + + +global logger +logger = Log() +``` + +### 配置模块 +```python +# Config.py +import configparser +import sys + + +global config + +config = configparser.ConfigParser() +config.sections() +config.read("setting.ini") +``` + +### 配置文件 +``` +[LOG] +LEVEL = INFO //日志等级 +LOG_PATH = ./FundCrawler/logs //日志目录 + +[MYSQL] +host = 127.0.0.1 //MYSQL服务器ip +PORT = 20137 //MYSQL服务器端口 +USERNAME = username +PASSWORD = password +DATA_BASE_NAME = Fund +``` + +### 爬虫主程序 + +```python +# PinyinDataCrawler.py +import random +import requests +import json +import pymysql +from Logger import logger +from Config import config + +header_str = '''Host:hanyu.baidu.com +Connection:keep-alive +sec-ch-ua:"Chromium";v="92", " Not A;Brand";v="99", "Microsoft Edge";v="92" +Accept:application/json, text/javascript, */*; q=0.01 +X-Requested-With:XMLHttpRequest +sec-ch-ua-mobile:?0 +User-Agent:Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/92.0.4515.131 Safari/537.36 Edg/92.0.902.67 +Sec-Fetch-Site:same-origin +Sec-Fetch-Mode:cors +Sec-Fetch-Dest:empty +Referer:https://hanyu.baidu.com/s?wd=%E5%8F%B7%E8%AF%8D%E7%BB%84&from=poem +Accept-Encoding:gzip, deflate, br +Accept-Language:zh-CN,zh;q=0.9,en;q=0.8,en-US;q=0.7 +Cookie:填写自己的Cookie''' + + +params_str = '''wd=%E5%8F%B7%E8%AF%8D%E7%BB%84 +from=poem +pn=1 +_=1628826451938''' + +class PinyinDataCrawler: + homographWeightDict = dict() + def __init__(self): + self.conn = self.getConnection() + self.character_list = self.getAllCharacters() + f = open("./data/luna_pinyin.dict.yaml", "r", encoding='utf8') + for line in f.readlines(): + datas = line.strip().split('\t') + if len(datas) == 3: + word = datas[0] + pinyin = datas[1] + weight = datas[2] + if word not in self.homographWeightDict: + self.homographWeightDict[word] = dict() + if pinyin not in self.homographWeightDict[word]: + self.homographWeightDict[word][pinyin] = dict() + self.homographWeightDict[word][pinyin] = weight + + def getHomograph(self, word="不"): + return self.homographWeightDict.get(word, dict()) + + # 把所有的多音字进行识别 + def splitHomograph(self, path='./Clover四叶草拼音', newPath='./Clover四叶草拼音new'): + if not os.path.exists(newPath): + os.mkdir(f'{newPath}') + + for file_now in os.listdir(path): + new_file_path = os.path.join(newPath, file_now) + curr_path = os.path.join(path, file_now) + new_file = open(new_file_path, 'w', encoding="utf-8") + if 'base' not in curr_path: + continue + for line in open(curr_path, encoding='utf-8'): + if "\t" in line: + keyword = line.split('\t')[0] + pinyin_old = line.split('\t')[1].strip() + count_str = line.split('\t')[-1].strip().replace(" ", '') + pinyinDict = self.getHomograph(keyword) + if len(pinyinDict) == 0: + new_file.write(line) + new_file.flush() + else: + currPinyins = sorted(pinyinDict.items(), key=lambda x: x[1], reverse=True) + for currPinyin in currPinyins: + try: + newLine = line.replace(pinyin_old, currPinyin[0]).replace(count_str, currPinyin[1]) + new_file.write(newLine) + new_file.flush() + except Exception as e: + print(e) + else: + new_file.write(line) + new_file.flush() + new_file.close() + + def format_header(self, header_str=header_str): + header = dict() + for line in header_str.split('\n'): + header[line.split(':')[0]] = ":".join(line.split(':')[1:]) + return header + + def format_params(self, params_str=params_str): + params = dict() + for line in params_str.split('\n'): + params[line.split('=')[0]] = line.split('=')[1] + return params + + def getPlainPinyin(self, sug_py): + splits = ['a', 'o', 'e', 'i', 'u', 'ü'] + shengdiao = '''a ā á ǎ à +o ō ó ǒ ò +e ē é ě è +i ī í ǐ ì +u ū ú ǔ ù +ü ǖ ǘ ǚ ǜ''' + shengdiaoToWord = dict() + for line in shengdiao.split("\n"): + datas = line.split(' ') + for curr in datas[1:]: + shengdiaoToWord[curr] = datas[0] + plain_pinyin = '' + for curr in sug_py: + if curr not in shengdiaoToWord: + plain_pinyin += curr + else: + plain_pinyin += shengdiaoToWord[curr] + return plain_pinyin + + def storeWord(self, word, currMean, word_index): + try: + pinyin = currMean.get("pinyin", "") + if len(pinyin) > 0: + pinyin = pinyin[0] + definition = currMean.get("definition", "") + if len(definition) > 0: + definition = "\n".join(definition).replace("'", '"') + if len(definition) > 4096: + definition = definition[:4096] + plain_pinyin = currMean.get("sug_py", "") + if len(plain_pinyin) > 0: + plain_pinyin = self.getPlainPinyin(pinyin) + pronunciation = currMean.get("tone_py", "") + if len(pronunciation) > 0: + pronunciation = pronunciation[0] + sql_str = f"insert into single_character_info values ('{pinyin}', '{word}', \ + '{plain_pinyin}', '{definition}', '{pronunciation}', {word_index})" + cursor = self.conn.cursor(pymysql.cursors.DictCursor) + cursor.execute(sql_str) + self.conn.commit() + except Exception as e: + if "Duplicate" not in f"{e}": + logger.error(f"word: {word}, pinyin: {pinyin}, error_info: {e}") + + def fixesDatas(self): + f = open('1.txt') + for line in f.readlines(): + word = line.replace(" ", "").split(":")[-3].split(",")[0] + logger.info(f"更新词组[{word}]的数据") + headers = self.format_header() + params = self.format_params() + params["pn"] = 1 + url = "https://hanyu.baidu.com/hanyu/ajax/search_list" + params['wd'] = word + while True: + try: + response = requests.get(url, params=params, headers=headers, timeout=(3.05, 10)) + break + except Exception as e: + logger.error(f"error_info: {e}") + if 'timed out' in f"{e}": + continue + datas = json.loads(response.text).get('ret_array', list()) + for currWordData in datas: + if 'mean_list' not in currWordData: + logger.warning(f"warning_info: {word} has not mean_list") + continue + currMeanList = currWordData["mean_list"] + try: + word = currWordData["name"][0] + for currMean in currMeanList: + self.storeWord(word, currMean, 0) + break + except Exception as e: + logger.error(f"error_info: {e}") + + def parserDatas(self, word, datas, word_index): + for currWordData in datas: + if 'mean_list' not in currWordData: + logger.warning(f"warning_info: {word} has not mean_list") + continue + currMeanList = currWordData["mean_list"] + word = currWordData["name"][0] + for currMean in currMeanList: + self.storeWord(word, currMean, word_index) + + def getConnection(self): + host = config["MYSQL"]["HOST"] + port = int(config["MYSQL"]["PORT"]) + db = config["MYSQL"]["DATA_BASE_NAME"] + user = config["MYSQL"]["USERNAME"] + password = config["MYSQL"]["PASSWORD"] + conn = pymysql.connect(host=host, port=port, db=db, user=user, password=password) + return conn + + def getCurrWordPageCount(self, url, params, headers): + pageCount = -1 + while pageCount == -1: + try: + response = requests.get(url, params=params, headers=headers, timeout=(3.05, 27)) + if len(response.text) == 0: + break + pageDatas = json.loads(response.text).get('extra', dict()) + pageCount = pageDatas.get("total-page", 0) + except Exception as e: + logger.error(f"getCurrWordPageCount, error_info: {e}") + time.sleep(random.randint(5, 10)) + + logger.info(f"getCurrWordPageCount, pageCount = {pageCount}") + return pageCount + + def crawlerExactPhrasePinyin(self, word="号", word_index=0, characters=None, phrase=True): + if characters is None and phrase is True: + characters = self.character_list + if characters is None: + characters = list() + headers = self.format_header() + params = self.format_params() + params["pn"] = 1 + url = "https://hanyu.baidu.com/hanyu/ajax/search_list" + if phrase: + params['wd'] = word + "词组" + else: + params['wd'] = word + pageCount = self.getCurrWordPageCount(url, params, headers) + currPageIndex = 0 + while currPageIndex < pageCount: + params["pn"] = currPageIndex + while params["pn"] == currPageIndex: + try: + logger.info(f"更新汉字[{word}]({word_index + 1} / {len(characters)})的词组数据(页数:{currPageIndex + 1} / {pageCount}) ") + response = requests.get(url, params=params, headers=headers, timeout=(3.05, 10)) + datas = json.loads(response.text).get('ret_array', list()) + self.parserDatas(word, datas, word_index) + currPageIndex += 1 + except Exception as e: + logger.error(f"更新汉字{word}出现错误, 错误信息: {e}") + time.sleep(random.randint(5, 9)) + time.sleep(random.randint(5, 100) * 0.01) + logger.info(f"更新汉字[{word}]的词组数据完成。") + + def getAllCharacters(self): + file_path = './data/clover.base.dict.yaml' + file = open(file_path, 'r', encoding="utf-8") + character_list = list() + for line in file.readlines(): + if '\t' in line: + word = line.split('\t')[0] + if word not in set(character_list): + character_list.append(word) + return character_list + + def getCurrCharacterStoreIndex(self): + sql_str = 'select * from single_character_info order by wordID desc' + cursor = self.conn.cursor(pymysql.cursors.DictCursor) + cursor.execute(sql_str) + data = cursor.fetchone() + logger.info(f'getCurrCharacterStoreIndex data = {data}') + if data is None: + return 0 + return data['wordID'] + + def crawlerPhraseDict(self): + characters = self.character_list + i = self.getCurrCharacterStoreIndex() + #i = 41318 + while i < len(characters): + word = characters[i] + logger.info(f'开始更新汉字[{word}]({i + 1}/{len(characters)})的数据...') + self.crawlerExactPhrasePinyin(word, i, characters) + i = i + 1 + logger.info(f'更新汉字[{word}]({i + 1}/{len(characters)})的数据完成！') + time.sleep(random.randint(1, 5) * 0.2) + +if __name__ == "__main__": + PinyinDataCrawler().crawlerPhraseDict() +``` +- 注意事项 +需要把header_str中的Cookie替换成你电脑浏览器生成的Cookie +### 运行程序 +- 把上述所有模块放在一个文件夹下，字典放在data目录下，进行项目目录运行下面代码 +```bash +python PinyinDataCrawler.py +``` +- 数据量比较多，大概有35W的数据，需要抓取2-3天左右，建议后台运行 +- 项目的源码：https://github.com/SivanLaai/BaiduPinyinCrawler +### 运行日志（部分展示） + +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218173958.png) + +### 数据库存储（部分展示） + +![image.png](https://cdn.statically.io/gh/SivanLaai/image-store-rep@master/note/20231218174038.png) \ No newline at end of file