<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[ab0902cd的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/ab0902cd</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; ab0902cd]]></copyright><item><title><![CDATA[nvidia docker安装]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/135476255</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/135476255</guid><author>ab0902cd</author><pubDate>Tue, 09 Jan 2024 11:32:32 +0800</pubDate><description><![CDATA[2. 安装nvidia docker。1.安装docker。3. gpu打进容器。]]></description><category></category></item><item><title><![CDATA[huggingface 下载大模型]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/134207035</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/134207035</guid><author>ab0902cd</author><pubDate>Fri, 03 Nov 2023 17:04:50 +0800</pubDate><description><![CDATA[大模型下载方法]]></description><category></category></item><item><title><![CDATA[ubutnu qt.qpa.xcb: could not connect to display 1]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/131705510</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/131705510</guid><author>ab0902cd</author><pubDate>Thu, 13 Jul 2023 16:23:14 +0800</pubDate><description><![CDATA[解决办法：export DISPLAY：=1。]]></description><category></category></item><item><title><![CDATA[堆排序python]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/128485073</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/128485073</guid><author>ab0902cd</author><pubDate>Thu, 29 Dec 2022 17:25:29 +0800</pubDate><description><![CDATA[堆排序]]></description><category></category></item><item><title><![CDATA[分组卷积与dw卷积]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/128469146</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/128469146</guid><author>ab0902cd</author><pubDate>Wed, 28 Dec 2022 15:33:54 +0800</pubDate><description><![CDATA[分组卷积、可分离卷积]]></description><category></category></item><item><title><![CDATA[免密登录跳板机和linux服务器]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/127010075</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/127010075</guid><author>ab0902cd</author><pubDate>Fri, 23 Sep 2022 14:34:45 +0800</pubDate><description><![CDATA[scp免密登录跳板机]]></description><category></category></item><item><title><![CDATA[Linux命令之sed批量替换字符串]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/126833936</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/126833936</guid><author>ab0902cd</author><pubDate>Tue, 13 Sep 2022 15:11:21 +0800</pubDate><description><![CDATA[批量替换字符串]]></description><category></category></item><item><title><![CDATA[给图片添加透明水印]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/126492382</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/126492382</guid><author>ab0902cd</author><pubDate>Tue, 23 Aug 2022 20:38:51 +0800</pubDate><description><![CDATA[透明水印]]></description><category></category></item><item><title><![CDATA[kaldi 源码安装]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/125783366</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/125783366</guid><author>ab0902cd</author><pubDate>Thu, 14 Jul 2022 14:17:39 +0800</pubDate><description><![CDATA[kaldi安装]]></description><category></category></item><item><title><![CDATA[GCC9.5离线安装]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/125541985</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/125541985</guid><author>ab0902cd</author><pubDate>Thu, 30 Jun 2022 15:50:10 +0800</pubDate><description><![CDATA[gcc]]></description><category></category></item><item><title><![CDATA[文本特征向量抽取及PCA降维]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/125104190</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/125104190</guid><author>ab0902cd</author><pubDate>Thu, 02 Jun 2022 17:19:12 +0800</pubDate><description><![CDATA[降维]]></description><category></category></item><item><title><![CDATA[多标签分类与binary_cross_entropy_with_logits]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/124413363</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/124413363</guid><author>ab0902cd</author><pubDate>Mon, 25 Apr 2022 21:14:48 +0800</pubDate><description><![CDATA[1. binary_cross_entropy_with_logits可用于多标签分类

torch.nn.functional.binary_cross_entropy_with_logits等价于torch.nn.BCEWithLogitsLoss

torch.nn.BCELoss+torch.nn.Sigmoid 等价于torch.nn.BCEWithLogitsLoss



在pytorch中torch.nn.functional.binary_cross_entropy_with_log...]]></description><category></category></item><item><title><![CDATA[tensorflow获取动态shape]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/123872696</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/123872696</guid><author>ab0902cd</author><pubDate>Thu, 31 Mar 2022 15:16:29 +0800</pubDate><description><![CDATA[tf.shape(a)和a.get_shape()比较

相同点：都可以得到tensor a的尺寸

不同点：tf.shape()中a 数据的类型可以是tensor, list, array

　　　　a.get_shape()中a的数据类型只能是tensor,且返回的是一个元组（tuple）

如果需要根据上一层的动态shape计算当前层的shape，该如何做呢

x=tf.placeholder(tf.float32, shape=[None, 227,227,3] )

但在运行的时候想知道None到]]></description><category></category></item><item><title><![CDATA[GCC安装]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/123850835</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/123850835</guid><author>ab0902cd</author><pubDate>Wed, 30 Mar 2022 17:27:22 +0800</pubDate><description><![CDATA[1.下载GCC安装包，下载地址如下，选择需要的安装版本：


https://mirrors.tuna.tsinghua.edu.cn/gnu/gcc/

2.解压配置

 进入解压目录执行：./configure

可能会遇到下面的问题：

configure: error: Building GCC requires GMP 4.2+, MPFR 2.4.0+ and MPC 0.8.0+.
Try the --with-gmp, --with-mpfr and/or --with-mpc optio.]]></description><category></category></item><item><title><![CDATA[scp通过跳板机传输文件]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/122875364</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/122875364</guid><author>ab0902cd</author><pubDate>Fri, 11 Feb 2022 10:44:22 +0800</pubDate><description><![CDATA[ssh 通过代理机跳转内网

proxyjump代理登录跳板机

使用proxyjump时，可能会遇到连接中转服务器后报错：
channel 0: open failed: administratively prohibited: open failed
stdio forwarding failed
kex_exchange_identification: Connection closed by remote host
这是因为没有指定端口号，需要指定端口号

还有可能遇到下面的问题：

ssh: Co]]></description><category></category></item><item><title><![CDATA[onnx重写输入和输出的维度]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/120200557</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/120200557</guid><author>ab0902cd</author><pubDate>Thu, 09 Sep 2021 14:28:24 +0800</pubDate><description><![CDATA[onnx模型输入是静态的，比如是1x3x960x960，但是想把输入改成动态输入，相应的输出也得改成动态，以下代码可以修改onnx模型的维度：


import onnx
import onnx.checker
import onnx.utils
from onnx.tools import update_model_dims
 
 
model = onnx.load('infer_rec.onnx')
# 此处可以理解为获得了一个维度 “引用”，通过该 “引用“可以修改其对应的维度           ]]></description><category></category></item><item><title><![CDATA[android cmake踩坑记录]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/115699623</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/115699623</guid><author>ab0902cd</author><pubDate>Wed, 14 Apr 2021 15:42:53 +0800</pubDate><description><![CDATA[ａｎｄｒｏｉｄ　ｓｔｕｄｉｏ最终正常编译ｓｏ库的环境配置如下：

ａｎｄｒｏｉ　ｓｔｕｄｉｏ　版本：３.５

ｇｒａｄｌｅ　ｖｅｒｓｉｏｎ　：5.6.3

ｇｒａｄｌｅ　plugin　：３.４.１

cmake配置：


cmake {
    path "src/main/cpp/CMakeLists.txt"
    version "3.10.2"
}

ｎｄｋ：android-ndk-r16b

问题1：

Execution failed for task ':app:generateJsonMo]]></description><category></category></item><item><title><![CDATA[感受野的计算]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/109893698</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/109893698</guid><author>ab0902cd</author><pubDate>Sat, 21 Nov 2020 15:03:58 +0800</pubDate><description><![CDATA[感受野的计算

感受野（Receptive Field），指的是神经网络中神经元“看到的”输入区域，在卷积神经网络中，feature map上某个元素的计算受输入图像上某个区域的影响，这个区域即该元素的感受野。

卷积神经网络中，越深层的神经元看到的输入区域越大，如下图所示，kernel size 均为3×33×3，stride均为1，绿色标记的是Layer2Layer2每个神经元看到的区域，黄色标记的是Layer3Layer3看到的区域，具体地，Layer2Layer2每个神经元可看到Layer1La..]]></description><category></category></item><item><title><![CDATA[C++多线程编程#pragma omp parallel]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/108770396</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/108770396</guid><author>ab0902cd</author><pubDate>Thu, 24 Sep 2020 11:52:26 +0800</pubDate><description><![CDATA[通常创建线程通过pthread_create来进行线程创建

创建线程

下面的程序，我们可以用它来创建一个 POSIX 线程：


#include &lt;pthread.h&gt;
pthread_create (thread, attr, start_routine, arg) 

在这里，pthread_create创建一个新的线程，并让它可执行。下面是关于参数的说明：

参数
			描述
		thread
			指向线程标识符指针。
		attr
			一个不透明的属性对象，可以被用来设置线程.]]></description><category></category></item><item><title><![CDATA[JNI如何避免内存泄露]]></title><link>https://blog.youkuaiyun.com/ab0902cd/article/details/108769716</link><guid>https://blog.youkuaiyun.com/ab0902cd/article/details/108769716</guid><author>ab0902cd</author><pubDate>Thu, 24 Sep 2020 11:19:07 +0800</pubDate><description><![CDATA[内存泄露主要是申请后没有及时释放导致的，谁申请谁释放是c++必须遵循的重要原则，malloc申请一般通过free进行释放，new申请的内存一般通过delete进行释放，

如果不进行释放，内存会逐渐升高，最终无法申请到内存而导致内存泄露

1. 哪些需要手动释放？


	不要手动释放（基本类型）： jint , jlong , jchar
	
	
	需要手动释放（引用类型，数组家族）： jstring，jobject ，jobjectArray，jintArray ，jclass ，jmethodID
	
]]></description><category></category></item></channel></rss>