<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[mmm305658979的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/mmm305658979</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; mmm305658979]]></copyright><item><title><![CDATA[ssh免密快捷连接简单配置实现]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/126268261</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/126268261</guid><author>mmm305658979</author><pubDate>Wed, 10 Aug 2022 15:59:08 +0800</pubDate><description><![CDATA[ssh快捷无密连接配置]]></description><category></category></item><item><title><![CDATA[PyTorch .pth 权重文件名称不一致]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/122668195</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/122668195</guid><author>mmm305658979</author><pubDate>Mon, 24 Jan 2022 15:19:44 +0800</pubDate><description><![CDATA[比如初始权重key为module.fnet.conv1.weight，但是用另一种方式训练的模型不需要module，key名称把这些拿掉就行了
# Performs input mapping such that stage 0 model can be loaded
if os.path.exists(args.model):
    prop_saved = torch.load(args.model)
    from collections import OrderedDict
    prop_s]]></description><category></category></item><item><title><![CDATA[电脑驱动正常蓝牙图标不见了]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/122628963</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/122628963</guid><author>mmm305658979</author><pubDate>Fri, 21 Jan 2022 20:36:27 +0800</pubDate><description><![CDATA[拔掉所有外接设备和电源。
关机。
长按开机20s，开机。

]]></description><category></category></item><item><title><![CDATA[Ubuntu ~/.bashrc 配置 for cuda torch]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/122426078</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/122426078</guid><author>mmm305658979</author><pubDate>Tue, 11 Jan 2022 10:35:01 +0800</pubDate><description><![CDATA[Ubuntu ~/.bashrc 配置 for cuda torch
convert xxx to your name
# ~/.bashrc: executed by bash(1) for non-login shells.
# see /usr/share/doc/bash/examples/startup-files (in the package bash-doc)
# for examples

# If not running interactively, don't do anything
]]></description><category></category></item><item><title><![CDATA[Ubuntu 安装 Github Desk]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/121952511</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/121952511</guid><author>mmm305658979</author><pubDate>Wed, 15 Dec 2021 14:52:27 +0800</pubDate><description><![CDATA[sudo wget https://github.com/shiftkey/desktop/releases/download/release-2.9.3-linux3/GitHubDesktop-linux-2.9.3-linux3.deb

# double click to install


]]></description><category></category></item><item><title><![CDATA[PIL Image P和L模式]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/121869673</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/121869673</guid><author>mmm305658979</author><pubDate>Sat, 11 Dec 2021 10:25:45 +0800</pubDate><description><![CDATA[通常RGB图像包含三个颜色通道，这些通道组合可以生成不同颜色。
【P mode】但是对于Mask而言，每个颜色通常表示一个物体，而物体的种类是有限的，因此可以用P模式，P模式需要预定义好多个颜色的模板，_palette= [0, 0, 0, 128, 0, 0, 0, 128,…]，因此只需要一个通道存储index即可，（index可以从模板中取出对应的颜色），这样表达颜色个数上限是255但是足够了，而且只占用1/3数据存储空间。
【L mode】同样只存储一个通道数据，但是代表的是灰度图的光照强度，而不是]]></description><category></category></item><item><title><![CDATA[python 下载google文件]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/121471460</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/121471460</guid><author>mmm305658979</author><pubDate>Mon, 22 Nov 2021 14:48:02 +0800</pubDate><description><![CDATA[import godown

gdown.download('https://drive.google.com/uc?id=1-QrceIl5sUNTKz7Iq0UsWC6NLZq7girr', output='../YouTube2018/valid.zip', quiet=False)



]]></description><category></category></item><item><title><![CDATA[Torch logs]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/121320473</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/121320473</guid><author>mmm305658979</author><pubDate>Sun, 14 Nov 2021 17:28:55 +0800</pubDate><description><![CDATA[各种pytorch小知识]]></description><category></category></item><item><title><![CDATA[Latex 各种tips]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/115121959</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/115121959</guid><author>mmm305658979</author><pubDate>Tue, 23 Mar 2021 14:21:30 +0800</pubDate><description><![CDATA[慢慢更新

字体大小
\tiny
\scriptsize
\footnotesize
\small
\normalsize
\large
\Large
\LARGE
\huge
\Huge

]]></description><category></category></item><item><title><![CDATA[torch tensor 改变尺寸操作]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/113069747</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/113069747</guid><author>mmm305658979</author><pubDate>Sun, 24 Jan 2021 11:08:21 +0800</pubDate><description><![CDATA[import torch

tensor = torch.rand(2,3,224,224)

tensor.is_contiguous()  # must make sure is contiguous

tensor.unsqueeze(0).shape  # torch.Size([1, 2, 3, 224, 224])
tensor.squeeze(0).shape  # Only cut the dimention if it is 1

tensor.view(-1,244,244).shape]]></description><category></category></item><item><title><![CDATA[熵，交叉熵，相对熵与KL散度，JS散度。 直白易懂]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/112623285</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/112623285</guid><author>mmm305658979</author><pubDate>Thu, 14 Jan 2021 18:09:50 +0800</pubDate><description><![CDATA[一个简单的理解熵交叉熵相对熵 AKA KL散度JS散度
熵
随机事件的不确定性度量，也是衡量分布内样本所有要的信息编码长度，熵越大，不确定性越大。
H(P)=ΣP(x)log1P(x)
H(P) = ΣP(x)log\frac{1}{P(x)}
H(P)=ΣP(x)logP(x)1​
交叉熵
用其他分布q衡量分布p内样本所需要的信息编码长度。H(P,Q) &gt;= H§
H(P,Q)=ΣP(x)log1Q(x)
H(P,Q) = ΣP(x)log\frac{1}{Q(x)}
H(P,Q)=ΣP(x)log]]></description><category></category></item><item><title><![CDATA[Ubuntu 多用户分别使用不同版本CUDA (2021亲测可行)]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/112395589</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/112395589</guid><author>mmm305658979</author><pubDate>Sat, 09 Jan 2021 17:47:42 +0800</pubDate><description><![CDATA[参考链接：

https://blog.youkuaiyun.com/qq_41202069/article/details/108286319

https://www.cnblogs.com/yhjoker/p/10972795.html





1. 下载想要的CUDA和对应版本CUDNN

2. 如下图安装CUDA

* 额外版本CUDA不要安装显卡驱动，不要添加软链接！

* 路径可以使用自己目录 /home/xxx/cuda/ e.t.c.


#..一堆协议说明...
#直接按q退出协议说明.
ac..]]></description><category></category></item><item><title><![CDATA[Ubuntu / Linux 各种操作指南]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/108003860</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/108003860</guid><author>mmm305658979</author><pubDate>Fri, 14 Aug 2020 14:28:07 +0800</pubDate><description><![CDATA[Ubuntu 各种操作指南环境配置文件修改hosts文件并且重启安装提示缺少依赖包软件列表和包更新清理开root权限卸载包修改文件夹所属组和组名 （解决root创建的文件其他用户无法操作问题）定时任务恢复误删除的系统文件文件和文件夹权限修改AnacondaPytorchGPU监控GPU对应任务查找 *对应gpu索引, 适用于nvidia-smi找不到任务安装本地文件Ubuntu安装显卡驱动 （建议安装recommend的驱动）
环境配置文件
/etc/profile
~/.bashrc
~/.condarc]]></description><category></category></item><item><title><![CDATA[Docker使用]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/107933150</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/107933150</guid><author>mmm305658979</author><pubDate>Tue, 11 Aug 2020 13:07:51 +0800</pubDate><description><![CDATA[这里写自定义目录标题Docker使用下载查看已有镜像
Docker使用
下载
docker pull pytorch/pytorch 下载docker镜像到本地
查看已有镜像
docker images

]]></description><category></category></item><item><title><![CDATA[Pytorch TensorboardX可视化]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/107484761</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/107484761</guid><author>mmm305658979</author><pubDate>Tue, 21 Jul 2020 11:28:18 +0800</pubDate><description><![CDATA[这里写自定义目录标题Pytorch TensorboardX可视化安装代码嵌入本地查看服务器日志可视化
Pytorch TensorboardX可视化
安装
pip install tensorboardX
pip install tensorboard
代码嵌入
from torch.utils.tensorboard import SummaryWriter
w = SummaryWriter(‘runs/’)
w.add_scalar('loss", loss, epoch)
本地查看服务器日志可视化]]></description><category></category></item><item><title><![CDATA[检查pytorch是否都可导]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/107231185</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/107231185</guid><author>mmm305658979</author><pubDate>Thu, 09 Jul 2020 16:10:04 +0800</pubDate><description><![CDATA[1.from torch.autograd import gradcheck
2. 如果计算图有一块不可导会报错

3. torch.requires_grad检查
]]></description><category></category></item><item><title><![CDATA[LeetCode Python]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/105465757</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/105465757</guid><author>mmm305658979</author><pubDate>Sun, 12 Apr 2020 10:50:34 +0800</pubDate><description><![CDATA[[LeetCode] 3. Longest Substring Without Repeating Characters 最长无重复字符的子串


Given a string, find the length of the longest substring without repeating characters.

Example 1:

Input: "abcabcbb"
O...]]></description><category></category></item><item><title><![CDATA[Detectron2 Transforms]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/105070817</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/105070817</guid><author>mmm305658979</author><pubDate>Tue, 24 Mar 2020 15:04:27 +0800</pubDate><description><![CDATA[fvcore/transforms/transform.py

detectron2/data/transforms/transform.py + PIL

提供底层实现。

__all__ = [
  "RandomBrightness",
  "RandomContrast",
  "RandomCrop",
  "RandomExtent",
  "RandomFlip...]]></description><category></category></item><item><title><![CDATA[Pytorch Detectron2 图像预处理]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/105060809</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/105060809</guid><author>mmm305658979</author><pubDate>Mon, 23 Mar 2020 23:14:25 +0800</pubDate><description><![CDATA[detectron2/data/detection_utils.py

detectron2/data/transforms.py




from detectron2.config import get_cfg
from detectron2.data import detection_utils
from detectron2.data import transforms as T

#...]]></description><category></category></item><item><title><![CDATA[Detectron2 Pytorch sampler 采样器]]></title><link>https://blog.youkuaiyun.com/mmm305658979/article/details/105046648</link><guid>https://blog.youkuaiyun.com/mmm305658979/article/details/105046648</guid><author>mmm305658979</author><pubDate>Mon, 23 Mar 2020 14:06:48 +0800</pubDate><description><![CDATA[pytorch_cuda10.2_python3.6/lib/python3.6/site-packages/torch/utils/data/sampler.py

sampler是最底层的类，像sequential_sampler都继承了此类。

SequentialSampler 按顺序进行采样，


from torch.utils.data.sampler import Sequent...]]></description><category></category></item></channel></rss>