<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[Gentleman的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/captainAAAjohn</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; captainAAAjohn]]></copyright><item><title><![CDATA[n-gram的理解与实现]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/125251047</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/125251047</guid><author>captainAAAjohn</author><pubDate>Sun, 12 Jun 2022 22:49:11 +0800</pubDate><description><![CDATA[n-gram 简单介绍与实现]]></description><category></category></item><item><title><![CDATA[BERT预训练]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/125040780</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/125040780</guid><author>captainAAAjohn</author><pubDate>Mon, 30 May 2022 09:35:12 +0800</pubDate><description><![CDATA[重新复现一下bert
项目地址: https://github.com/stevezhangz/BERT_PRETRAIN_PYTORCH\url{https://github.com/stevezhangz/BERT_PRETRAIN_PYTORCH}https://github.com/stevezhangz/BERT_PRETRAIN_PYTORCH
操作方法
git clone https://github.com/stevezhangz/BERT_PRETRAIN_PYTORCH.git
cd B]]></description><category></category></item><item><title><![CDATA[【牛客网】刷题记录]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/124604040</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/124604040</guid><author>captainAAAjohn</author><pubDate>Mon, 09 May 2022 12:41:37 +0800</pubDate><description><![CDATA[不断安更新
刷题顺序参考B站教程https://space.bilibili.com/38725189?spm_id_from=333.788.b_765f7570696e666f.1\url {https://space.bilibili.com/38725189?spm_id_from=333.788.b_765f7570696e666f.1}https://space.bilibili.com/38725189?spm_id_from=333.788.b_765f7570696e666f.1
注：不同]]></description><category></category></item><item><title><![CDATA[【解决|ubuntu】d4rl 安装 踩坑记录]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/123024952</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/123024952</guid><author>captainAAAjohn</author><pubDate>Sat, 19 Feb 2022 22:50:44 +0800</pubDate><description><![CDATA[安装这玩意儿真心累啊，所以写一篇帖子去介绍下如何去搭建d4rl环境，希望可以推动科研的进步。
mujoco210是最新版本的mujoco，不需要key也不需要copy license。
这里建议使用linux系统，因为win10下不好操作。
我使用的是ubuntu20.04。
本文分以下几个环节介绍：(1) 如何搭建mujoco210 (2) 如何安装dm_control(3) 如何安装d4rl
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
一、搭建]]></description><category></category></item><item><title><![CDATA[ERROR: No matching distribution found for execjs]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/122556046</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/122556046</guid><author>captainAAAjohn</author><pubDate>Tue, 18 Jan 2022 11:17:54 +0800</pubDate><description><![CDATA[有的时候包的名字和下载名称对不上号就很棘手
pip install PyExecJS 

参考https://blog.youkuaiyun.com/lanluyug/article/details/84031474\url{https://blog.youkuaiyun.com/lanluyug/article/details/84031474}https://blog.youkuaiyun.com/lanluyug/article/details/84031474

...]]></description><category></category></item><item><title><![CDATA[[python]jinjia2安装不上的解决方法]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121883496</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121883496</guid><author>captainAAAjohn</author><pubDate>Sun, 12 Dec 2021 09:51:22 +0800</pubDate><description><![CDATA[网上的解答方法五花八门
亲测有效的方法:
pip3 install jinjia2 -i https://pypi.tuna.tsinghua.edu.cn/simple 


]]></description><category></category></item><item><title><![CDATA[Mujoco如何编写自己的环境]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121872165</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121872165</guid><author>captainAAAjohn</author><pubDate>Sun, 12 Dec 2021 09:46:29 +0800</pubDate><description><![CDATA[摸索一下如何建立mojoco环境，并记录一下某些参数的性质。
首先附加一些可能会用到的网址：

mujoco_py官方文档：https://github.com/openai/mujoco-py\url{https://github.com/openai/mujoco-py}https://github.com/openai/mujoco-py
mujoco xml中的参数解读:https://mujoco.readthedocs.io/en/latest/XMLreference.html\url{htt]]></description><category></category></item><item><title><![CDATA[win10安装mujoco]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121839644</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121839644</guid><author>captainAAAjohn</author><pubDate>Thu, 09 Dec 2021 18:51:13 +0800</pubDate><description><![CDATA[转自https://blog.youkuaiyun.com/wavehaha/article/details/120216301\url{https://blog.youkuaiyun.com/wavehaha/article/details/120216301}https://blog.youkuaiyun.com/wavehaha/article/details/120216301
还需要进一步实践，方可弄明白是否有效。

]]></description><category></category></item><item><title><![CDATA[关于hydra安装失败的解决方法]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121806923</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121806923</guid><author>captainAAAjohn</author><pubDate>Thu, 09 Dec 2021 09:55:35 +0800</pubDate><description><![CDATA[主要是安装包的名字不是hydra
pip install hydra-core

本文转自https://blog.youkuaiyun.com/weixin_43918682/article/details/121278570?utm_medium=distribute.pc_aggpage_search_result.none-task-blog-2~aggregatepage~first_rank_ecpm_v1~rank_v31_ecpm-1-121278570.pc_agg_new_rank&amp;utm_]]></description><category></category></item><item><title><![CDATA[解决Ubuntu pytcharm无法输入中文]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121711288</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/121711288</guid><author>captainAAAjohn</author><pubDate>Sat, 04 Dec 2021 09:26:08 +0800</pubDate><description><![CDATA[这篇博客解决的很完美
https://blog.youkuaiyun.com/m0_54032194/article/details/119081008?spm=1001.2101.3001.6650.1&amp;utm_medium=distribute.pc_relevant.none-task-blog-2%7Edefault%7ECTRLIST%7Edefault-1.showsourcetag&amp;depth_1-utm_source=distribute.pc_relevant.none-task-bl]]></description><category></category></item><item><title><![CDATA[Leetcode hot 100题之前50题刷题记录]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/120801091</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/120801091</guid><author>captainAAAjohn</author><pubDate>Sat, 30 Oct 2021 23:38:21 +0800</pubDate><description><![CDATA[计算两个链表中的数值之和
https://leetcode-cn.com/problems/add-two-numbers/submissions/
思路：初始化首尾两个指针，然后每一位相加，同时不能忘记进位。


]]></description><category></category></item><item><title><![CDATA[[硬件保护]让笔记本降温的方法]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/120155700</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/120155700</guid><author>captainAAAjohn</author><pubDate>Tue, 07 Sep 2021 13:15:31 +0800</pubDate><description><![CDATA[去年8月份入手了一台 i5 10400h+ RTX 2070 115w的笔记本，使用了一年有一件事情一直在困惑着我，笔记本高温是一个普遍现象还是个别现象。当我的笔记本一开启性能模式并运行深度学习或者游戏等吃硬件的程序，GPU和CPU温度飙升至87度，虽然这在网上不算最高的，但是我很好奇的是一些人可以把相同配置的本子温度压到80以下，究竟是什么原因造成了这种差距。接下来我将带来几方面的分析。
1.硅脂
我电脑的原配硅脂很垃圾，我估计是厂商使用了三块钱一桶的硅脂进行涂的。说实在的我不太懂他们的想法，比如7921]]></description><category></category></item><item><title><![CDATA[强化学习与数学]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/119509843</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/119509843</guid><author>captainAAAjohn</author><pubDate>Mon, 09 Aug 2021 22:50:46 +0800</pubDate><description><![CDATA[参考文献:
[1]Agent 57:Outperforming the Atari Human Benchmark[1]\mathcal{Agent\,57:Outperforming \,the \,Atari \,Human \,Benchmark }[1]Agent57:OutperformingtheAtariHumanBenchmark
[2]Renforcement Learning(Sutton)[2]\mathcal{Renforcement \, Learning(Sutton)}[2]R]]></description><category></category></item><item><title><![CDATA[机器学习之 监督学习]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118957707</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118957707</guid><author>captainAAAjohn</author><pubDate>Wed, 21 Jul 2021 11:53:28 +0800</pubDate><description><![CDATA[(不断更新)
机器学习可分为监督学习和五监督学习两大类，监督学习需要依赖于人为标注的数据，通过选择合适地目标函数令模型的预测不断地贴近监督信息，从而得到精准的预测。无监督学习，意在学习到数据中的潜在特征，并不涉及到监督数据。
监督学习
线性回归
define： 给定特征数据集XXX,给定标签YYY，通过MSEerrorMSE_{error}MSEerror​去减小模型的预测 Y^\hat{Y}Y^同标签YYY之间的差距。
推导(1)----------优化指标
(1)深度学习与机器学习框架
一些人说机器学习]]></description><category></category></item><item><title><![CDATA[简单的嵌入式人脸识别系统]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118583355</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118583355</guid><author>captainAAAjohn</author><pubDate>Thu, 08 Jul 2021 22:01:43 +0800</pubDate><description><![CDATA[今天基于卷积神经网络给一个简单的嵌入式人脸识别系统
思路如下：登记时，我给了10个容量，可以给需要登记的人一个标签，然后pattern选择‘train’即可实现人脸登记，按’q‘推出截屏。
使用时，选择’val‘截屏后判断你是哪一个人，opeartion可以是你自定义的函数，比如你想实现一个人脸识别开所程序，那么opearion只要是传入和pre之间有联系的值即可。
import cv2
from torchvision.models import resnet
import torch
from torc]]></description><category></category></item><item><title><![CDATA[常用矩阵求导]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118215413</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118215413</guid><author>captainAAAjohn</author><pubDate>Fri, 25 Jun 2021 19:18:53 +0800</pubDate><description><![CDATA[矩阵求导的本质：假设Y=ABY=ABY=AB,Y对A求导意味着Y中每一个元素对A进行求导。
我们由简单到复杂解释如何进行矩阵求导：
对于
x=[x1x2⋅⋅xn] y=f(x)
x= \begin{bmatrix}
  x_1 \\
   x_2 \\
   ·\\
   ·\\
   x_n
  \end{bmatrix} \,
  y=f(x)
x=⎣⎢⎢⎢⎢⎡​x1​x2​⋅⋅xn​​⎦⎥⎥⎥⎥⎤​y=f(x)
∇xy=[dydx1dydx2⋅⋅dydxn]\nabla_{x}y= \begin{bm]]></description><category></category></item><item><title><![CDATA[[彻底解决]CUDA error: an illegal memory access was encountered(CUDA错误 非法访问内存)]]></title><link>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118162508</link><guid>https://blog.youkuaiyun.com/captainAAAjohn/article/details/118162508</guid><author>captainAAAjohn</author><pubDate>Wed, 23 Jun 2021 21:04:21 +0800</pubDate><description><![CDATA[先说一下在网上看到的问题：
第一种可能你的程序涉及到并行计算，但你只有一张卡，因此只要将程序涉及到并行计算的部分改成单卡即可
找找有没有 torch.nn.DataParallel()

第二种一部分数据或者模型在cpu上，另外一部分在gpu上。
PS：第二种和第一种可以强行 os.environ["CUDA_VISIBLE_DEVICES"]为指定ID

第三种CUDNN版本不对？感觉这种可能比较小
然后说一下我是怎么解决的：
最关键的，也是我遇到的问题，这个错误没有表明和显存溢出存在着联系，因为显存溢出]]></description><category></category></item></channel></rss>