问了gpt几个适合孩子打字的网站，选择了typing club，看起来娱乐性不错，打算让果果试一试。主要培养孩子如下几点：
前几天网站的图片遭到了不法分子的攻击，怪我没有安全防范意识，bucket还是用公有读，配置COS Referer白名单即可。
The LLM model is to big to put into the device, if there any batter architecture to make the model small enough to put to the device and keep the precision and inference performance.
The LLM model embedding the word to very high dimension, in order to do better classification. But it will affected by the curse of dimensionality. So we need deep neural net work(more parameters) and more training data to avoid overfitting.
Also use attention mechanism to make them classifiable.
But each time put some words into the space, may be adding one new dimension only useful for related words, and for most other words will add one no useful dimension. And the whole vector space is sparse and need more training data and complex model(more parameters).
We can use the idea of mixture of experts, they form one graph, and
some experts are independent, but some has relations between each other,
can be connected by edges. So the problem becomes make each experts'
dimension as small as possible, and the duplication features in
different experts as small as possible. We convert one high dimension
sparse LLM to many row dimension dense experts. Just like the symbol
table design in compiler, we can use one big flatten symbol table, but
the scope level and scope name in each scope
are same and can be eliminated, so traditional implementation of the
symbol table use the chain of scopes hash table, make a more compact
memory.
A graph
Take the compute system as the whole world. We can split experts to
Once a word file system coming, will be both added to
operating system and databse, and if a word
computer arrived, it will be put to the
common knowledge. And all experts depend on the
common knowledge.
All the experts can be version controlled, and can evolve,
refactoring. Like add more words, add dimension, remove words to other
experts, extract more experts. Just like the package manage system, all
experts are packages.
The query convert to tokens, and send to the related experts,
exploit method use dependent experts to combine the
results, exploare method to query other experts to find
results too. The results contian all the experts, the sub graph forms a
path. The path can be cached.
Support real time input sentences. The tokenizer first to
classification, and put to different experts, batch will trigger insert.
Insert data to the experts, and update the path.
During the period 1904 to1947, the vacuum tube was the electronic device of interest and development. In 1904, the vacuum-tube diode was introduced by J. A. Fleming. Shortly thereafter, in 1906, Lee De Forest added a third element, called the control grid, to the vacuum diode, resulting in the first amplifier, the triode. In the following years, radio and television provided great stimulation to the tube industry. Production rose from about 1 million tubes in 1922 to about 100 million in 1937. In the early 1930s the four- element tetrode and the five-element pentode gained prominence in the electron-tube industry. In the years to follow, the industry became one of primary importance, and rapid advances were made in design, manufacturing techniques, high-power and high-frequency applications, and miniaturization.
三极管的发明，和工业界的需求有密切联系。radio and television provided great stimulation to the tube industry.
看现在的AI需求，对于新计算机体系结构也会有新的需求，比如memory
system的演进，可能会有比较革命性的变革，比如pim。
On December 23, 1947, however, the electronics industry was to experience the advent of a completely new direction of interest and development. It was on the afternoon of this day that Dr. S. William Shockley, Walter H. Brattain, and John Bardeen demonstrated the amplifying action of the first transistor at the Bell Telephone Laboratories. The original transistor (a point-contact transistor) is shown in Fig. 2. The advantages of this three-terminal solid-state device over the tube were immediately obvious: It was smaller and lightweight; it had no heater requirement or heater loss; it had a rugged construction; it was more efficient since less power was absorbed by the device itself; it was instantly available for use, requiring no warm-up period; and lower operating voltages were possible. Note that this chapter is our first discussion of devices with three or more terminals. You will find that all amplifiers (devices that increase the voltage, current, or power level) have at least three terminals, with one controlling the flow or potential between the other two.
为什么这些创新都是来自美国，Bell实验室，中国到现在为止，始终没有这样原创性的工作，抄永远是跟在别人屁股后面的。这个原型满足了一些基本特性需求：
在研发过程中，应该是有这些目标牵引进行设计探索的。所以从底层原理出发，解决工业界比较强烈的需求，可以做出一些原创性的事情。
The circuit consists of a resistor
The output voltage is also constrained by the supply voltages
根据下图，计算增益和
Solution:
The output voltage
要计算amplifier的linear region，输入幅值需要小于，
黄色为输入，左下角一个格子为1v，而蓝色一个格子为5v。
发现输出的顶部已经不再是正弦波。和理论计算比较一致。
运算放大器需要
所需要的元件：
参考文章：
https://blog.csdn.net/qq_41836400/article/details/98345011
include/apue.h在include/apue.h增加 1
stdio/buf.c打开apue.3e/stdio/buf.c，搜索_LP64，大致在（文件的89～93行），将其删除
1
2
3
4
5
将is_unbuffered,
is_linebuffered和buffer_size替换如下
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27int
is_unbuffered(FILE *fp)
{
return(fp->_flags & _IONBF);
}
int
is_linebuffered(FILE *fp)
{
return(fp->_flags & _IOLBF);
}
int
buffer_size(FILE *fp)
{
return(fp->_IO_buf_end - fp->_IO_buf_base);
return(BUFSIZ); /* just a guess */
}
1 | sudo cp ./include/apue.h /usr/include/ |
1 | cd |
add lines at the end of ~/.tmux.conf.local
1 | # force Vi mode |
add 185.199.108.133 raw.githubusercontent.com t
o/etc/hosts
1 | sudo apt install zsh |
install nvm
1 | curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.1/install.sh | bash |
npm源
1 | npm config set registry https://registry.npmmirror.com |
1 | npm install -g hexo-cli |
1 | cd blog |
1 | diff --git a/_config.yml b/_config.yml |
cat /etc/apt/sources.list
1 | deb http://mirrors.aliyun.com/raspbian/raspbian/ bullseye main non-free contrib rpi |
安装emacs依赖
1 | sudo apt install libxpm-dev libgif-dev libgnutls28-dev libgtk-3-dev libjpeg-dev libtiff-dev libncurses-dev libgccjit-10-dev |
1 | ./configure --with-native-compilation |
参考炬峰的视频，来搞个电台玩玩。
1 | git clone https://github.com/F5OEO/rpitx |
audio输入输出设置
验证输入输出
1 | arecord -f cd - | aplay -V sterio - |
修改testfmrds.sh脚本
1 | sudo arecord -f cd -D plughw:2,0 -c 2 - | sudo ./pifmrds -freq "$1" -audio - |
修改/boot/config.txt
1 | # Enable audio (loads snd_bcm2835) |
测试
1 | ./testfmrds.sh 434 |
build on the pi and dev machine
1 | cd ~/code/dep_ws/src |
for normal rplidar
1 | sudo apt install ros-foxy-rplidar-ros |