<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[遇事不决可问春风的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/walker_Tjj</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; walker_Tjj]]></copyright><item><title><![CDATA[RuntimeError: Expected object of scalar type Long but got scalar type Float for argument #2 ‘target‘]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/124392699</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/124392699</guid><author>walker_Tjj</author><pubDate>Sun, 24 Apr 2022 21:47:11 +0800</pubDate><description><![CDATA[pytorch框架模型报错：RuntimeError: Expected object of scalar type Long but got scalar type Float for argument #2 'target']]></description><category></category></item><item><title><![CDATA[The “freeze_support()“ line can be omitted if the program isnot going to be frozen to produce an]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/124234920</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/124234920</guid><author>walker_Tjj</author><pubDate>Wed, 20 Apr 2022 18:24:18 +0800</pubDate><description><![CDATA[The "freeze_support()" line can be omitted if the program
        is not going to be frozen to produce an executable.]]></description><category></category></item><item><title><![CDATA[Windows10 中使用nvidia-smi]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/124220223</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/124220223</guid><author>walker_Tjj</author><pubDate>Sat, 16 Apr 2022 21:27:02 +0800</pubDate><description><![CDATA[Windows 10 中使用nvidia-smi]]></description><category></category></item><item><title><![CDATA[模型训练时报错：RuntimeError: Unable to find a valid cuDNN algorithm to run convolution]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/123874520</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/123874520</guid><author>walker_Tjj</author><pubDate>Thu, 31 Mar 2022 16:01:56 +0800</pubDate><description><![CDATA[训练模型时报错，记录一下。这个错误是现存模型过大，GPU显存不足导致的错误，此时只需要将batch-size改小即可：128-&gt;64, 64--&gt;32等等。
]]></description><category></category></item><item><title><![CDATA[Win10系统Anaconda3创建虚拟环境，notebook 配置虚拟环境]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/118273300</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/118273300</guid><author>walker_Tjj</author><pubDate>Sun, 27 Jun 2021 15:27:39 +0800</pubDate><description><![CDATA[安装环境

系统：window10 - 64位
	软件：anaconda 4.4.0 - python3.6
1、 创建虚拟环境

1.1 默认安装虚拟环境

anaconda安装python3.6版本的虚拟环境的如下：



输入安装命令：


conda create -n py36 python=3.6


参数：

-n：会将虚拟环境安装在anaconda已安装路径默认的envs目录下。
	python=3.6：安装的虚拟环境为python3.6版本
备注：此命令只会在py36虚拟环境中安装少量的.]]></description><category></category></item><item><title><![CDATA[tensorflow.python.framework.errors_impl.UnknownError: 2 root error(s) found.]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/107195166</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/107195166</guid><author>walker_Tjj</author><pubDate>Wed, 08 Jul 2020 02:31:49 +0800</pubDate><description><![CDATA[在利用Tensorflow做后台，训练Keras代码时（当时版本为：TensorFlow1.14.0，Keras2.2.5，CUDA10.0，cudnn7.4.2）,爆出了错误如下：

tensorflow.python.framework.errors_impl.UnknownError: 2 root error(s) found.
 (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN fai.]]></description><category></category></item><item><title><![CDATA[tensorflow 模型量化（先占坑，随后写）]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/106953895</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/106953895</guid><author>walker_Tjj</author><pubDate>Fri, 26 Jun 2020 17:06:27 +0800</pubDate><description><![CDATA[传送门（需要翻墙）

官方指南：
https://www.tensorflow.org/lite/performance/post_training_quantization

Colab链接：
https://colab.research.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb
...]]></description><category></category></item><item><title><![CDATA[使用google Colaboratory的记录（免费K80 GPU）（二）]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/106960406</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/106960406</guid><author>walker_Tjj</author><pubDate>Fri, 26 Jun 2020 15:29:31 +0800</pubDate><description><![CDATA[Google Colaboratory搭建及使用教程（二） 

前言

上节讲到如何注册并使用Google Colaboratory运行测试我们的python程序。

传送：使用google Colaboratory的记录（免费K80 GPU）（一）--https://blog.youkuaiyun.com/walker_Tjj/article/details/106955426

但对于一个深度学习的工程项目来说，.ipynb文件只能运行简单功能程序，并不能满足我们的需求。大量...]]></description><category></category></item><item><title><![CDATA[使用google Colaboratory的记录（免费K80 GPU）（一）]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/106955426</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/106955426</guid><author>walker_Tjj</author><pubDate>Thu, 25 Jun 2020 17:34:23 +0800</pubDate><description><![CDATA[Google Colaboratory搭建及使用教程  

   接触过深度学习的都知道GPU对于我们来说意味着着什么（是金钱、是时间、是生命）。听说google提供的Colaboratory可以使用T4GPU，甚至P100GPU。本博主是历经重重险阻，登高、跳远，终于使用上了，本来一度欢欣鼓舞，但事实让我明白资本家的羊毛不是那么容易薅的。

一、前言

1、一切建立在“你懂的”的基础上

国内的网络肯定是使用不了的。想使用你首先需要一个谷歌账号（QQ邮...]]></description><category></category></item><item><title><![CDATA[Tensorflow所遇问题解决方案（持续更新）]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/106460283</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/106460283</guid><author>walker_Tjj</author><pubDate>Sun, 31 May 2020 20:22:49 +0800</pubDate><description><![CDATA[问题1：使用训练结束保存的logs文件，报错：


(base) E:\pythonprojects\Model_quantification\Model_quantification_mnist&gt;tensorboard --logdir==logdir
Traceback (most recent call last):
 File "c:\anaconda3\lib\runpy.py", line 193, in _run_module_as_main
  "__main__", mod_s...]]></description><category></category></item><item><title><![CDATA[tensorflow将张量增/减一个维度：tf.expand_dims()/tf.squeeze()]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/104600048</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/104600048</guid><author>walker_Tjj</author><pubDate>Sun, 01 Mar 2020 21:10:36 +0800</pubDate><description><![CDATA[1、TensorFlow中，想要维度增加一维，可以使用tf.expand_dims(input, dim, name=None)函数。


tf.expand_dims(
  input,
  axis=None,
  name=None,
  dim=None
)

参数：


	input是输入张量。
	
	
	axis是指定扩大输入张量形状的维度索引值。
	
	
	dim等同...]]></description><category></category></item><item><title><![CDATA[那些让人耳目一新的深度学习图像超分辨率重建方法]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/96315536</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/96315536</guid><author>walker_Tjj</author><pubDate>Wed, 17 Jul 2019 23:09:12 +0800</pubDate><description><![CDATA[本文章只记录（按时间），不分析网络结构、优劣。可能会有遗漏，见谅.

1. SRCNN --ECCV2014

   paper：Learning a Deep Convolutional Network for Image Super-Resolution

   source code：http://mmlab.ie.cuhk.edu.hk/projects/SRCNN.ht...]]></description><category></category></item><item><title><![CDATA[DeepLabv3+：语义分割领域的新高峰]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/80886907</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/80886907</guid><author>walker_Tjj</author><pubDate>Mon, 02 Jul 2018 17:05:58 +0800</pubDate><description><![CDATA[转载自：https://www.sohu.com/a/224351199_500659这是 PaperDaily 的第 44篇文章关于作者：陈泰红，小米高级算法工程师，研究方向为人脸检测识别，手势识别与跟踪。■ 论文 | Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation■ 链接 | h...]]></description><category></category></item><item><title><![CDATA[条件随机场(CRF)]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/80848374</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/80848374</guid><author>walker_Tjj</author><pubDate>Thu, 28 Jun 2018 20:01:49 +0800</pubDate><description><![CDATA[条件随机场(CRF) - 1 - 简介：https://www.2cto.com/kf/201605/512289.html条件随机场(CRF) - 2 - 定义和形式：https://www.2cto.com/kf/201605/512286.html]]></description><category></category></item><item><title><![CDATA[最优化问题解题流程]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/80845179</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/80845179</guid><author>walker_Tjj</author><pubDate>Thu, 28 Jun 2018 16:25:47 +0800</pubDate><description><![CDATA[]]></description><category></category></item><item><title><![CDATA[MatConvNet+Matlab2017a+CUDA8.0安装]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/79973500</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/79973500</guid><author>walker_Tjj</author><pubDate>Tue, 17 Apr 2018 13:28:48 +0800</pubDate><description><![CDATA[安装过程参照MatConvNet官网给出的步骤：http://www.vlfeat.org/matconvnet/install/1.安装CUDA8.0+cudnn6.0 参见之前的博客2.安装MatConvNet:下载地址：matconvnet-1.0-beta25.tar.gz安装步骤：我先安装了CPU版，然后安装了GPU版：CPU版：1.解压matconvnet到自己的目录。2.打开MatL...]]></description><category></category></item><item><title><![CDATA[win10下 CUDA9.0 + CuDNN7.0.5 安装回顾]]></title><link>https://blog.youkuaiyun.com/walker_Tjj/article/details/79966449</link><guid>https://blog.youkuaiyun.com/walker_Tjj/article/details/79966449</guid><author>walker_Tjj</author><pubDate>Mon, 16 Apr 2018 21:03:30 +0800</pubDate><description><![CDATA[入了深度学习的坑，需要搭建Tensorflow环境，虽然渣渣显卡，但是总比CPU来得快，果断选择GPU版。在网上找了很多资料，受益颇多。但是由于tensorflow最近更新了，CUDA8.0不能支持，得改用CUDA9.0。而网上资料对应的是老版本tensorflow，只支持CUDA8.0上，结果走了不少弯路。记下搭建过程，希望给大家一个参考。（老版本的搭建过程大家可以参考文末链接） 重要：请注意各...]]></description><category></category></item></channel></rss>