DEV Community: Chris Kay

Machine Learning in Rust; Is it worth learning it?

Chris Kay — Mon, 22 May 2023 14:05:43 +0000

As of 2023 what does the Rust language offer for ML and AI? How does it compare to the Giants ( e.g. Pytorch, Tensorflow libraries )? Most importantly, does it help you towards becoming a professional Ml engineer? These are the questions I had as a newcomer to the ML field and a programming language enthusiast. I tried to scratch the surface and report my findings so you can also have an idea of ML outside of Python, C++.

Is it hard?

Yes, especially if you come from dynamically-typed languages ( Python, Javascript ). Rust is a VERY strict/statically typed language, who benefits from this? Well actually Rust converts a lot of bugs of dynamically languages to actual errors, but do not have any high expectations for the first time. Rust also has many features high-level don't have: macros, ownership/borrowing, traits/implementations, Results ( error handling ) and probably other. Learning Rust is a challenge by itself, but if you have the balls, then is it worth it?

Is it worth it?

Yep, absolutely. You will learn not only AI by creating them from the ground up but also low-level programming, a skill not many master or even know. There is no Keras or torch.NN api that has all the layers you will never probably use, there is Linfa but is compares more to Scikit learn that Keras or torch.NN. If you want a Conv2d layer then you will have to build it your own. It is a feat to build a Deep Convolutional Neural Network in Rust that you will put in your Résumé. I will give you a pause to decide whether you want to embark on such journey and next I will talk about what APIs are available in Rust and if they are worth learning.

ML/AI APIs in RUST

In short here is a list of the some useful ML/AI crates in Rust

The Linfa and Smartcore crates are both similar to scikit-learn, ndarray provides multi-dimensional vectors ( tensors ), petgraph provides an API for graphs,ndarry_image helps you read images and convert them to RGB 2d arrays and rust-gpu allows you to use the gpu but I don't know if it is compatible for ML and the other crates. There are probably other libraries as well, but I didn't find any crate for Natural Language Processing ( NLP ) or Audio Processing. Please recommend any other crates if you find them helpful. Now the most important question is, what can you actually build without developing a Tensorflow for Rust library?

What can you create in a descent amount of time?

Let's start with what you will not create:

NLP AIs, don't expect to build an Alexa or Siri in Rust, the - technology still isn't there.
Audio Processing, no fake voices.
GANs, not these kind of guns you are thinking but Generative Adversarial Networks ( e.g. midjourney, dall-e )

Ok, now that you have some blurry expectations here is what you can probably create:

Regression models ( linfa, SmartCore )
Computer Vision Models ( cv )
Nearest Neighbor algorithms ( linfa, SmartCore )
Classification Models, there is no goo API you will have to create an API yourself ( your skills )
Supervised Models ( linfa, SmartCore )

There are some resources that you can follow but they intend to get your feet wet and not much else like this Youtube video that builds a neural network from scratch in Rust.

GPU support

Good luck :).

Learn ML in C++, it would be easier…

Learn C++ which is easier than Rust as it follows paradigms from popular languages and doesn't have many complex concepts, if you happen to not know every popular python ML library uses C++ in the back end which means that they have an API in C++. Tensorflow for example has Tensorflow Lite, pytorch also supports an API in c++ and opencv was built in C++. Moreover C++ is faster and will speed up your workflow.

Looking for something easier?

If you want to learn more about ML and you happen to know a little bit Web development I made a blog on how you can integrate Tensorflow with React.js. If you are just starting out then learn Pytorch be careful to not get stuck in tutorial hell though.

Create a Public File Sharing Server with NodeJS,ReactJS and Port Forwarding.

Chris Kay — Sat, 20 May 2023 06:33:26 +0000

Today We will Develop, Build and Publish a File Sharing Server on the Web anyone can join from Anywhere for Free. We will be using Nodejs and Reactjs for the development and Port Forwarding for Publishing our Site Internationally. If you don’t use this stack but find the project worthwhile doing then I you might want to check this out.

We will start with the server first. Firstly create a folder for the server and go in to it, after that initialize a node project. Use the y option to skip the interrogation. If you are unsure you have Nodejs on your system, check the version of node and if Nodejs is not installed go to https://nodejs.org/en and download the LTS version.



mkdir server
cd server

# Check for node version
node -v

# Initialize project
npm init -y

Install the project dependencies: express, express-fileupload, nodemon. Express will be used to create the web server and express-fileupload will be used for accessing the uploaded files.



npm i express express-fileupload

Add the main javascript file. For this example I will be naming it main.js
Add a script to start the server under the package.json file. You can also delete the test script. At the end this must be your package.json.



{
  "name": "server",
  "version": "1.0.0",
  "description": "",
  "main": "index.js",
  "scripts": {
    "start": "node main.js"
  },
  "keywords": [],
  "author": "",
  "license": "ISC"
}

Now we must write the script. I won’t go to the details but the script can be roughly separated in 3 parts: The Imports, the Boilerplate app statements and the routes. The fs.append code blocks are trackers for our website, they track the browser’s IP and route they are on



// Imports

const fs = require('fs')
const path = require('path')
const express = require('express')
const fileupload = require('express-fileupload')
const cors = require('cors')

// Boilerplate app statements

const app = express()

app.use(fileupload({
    useTempFiles: true,
    tempFileDir: "/tmp/"
}))
app.use(express.json())
app.use(cors({ origin: "*", }))

// helper functions

function padTo2Digits(num) {
    return num.toString().padStart(2, '0');
}

function formatDate(date) {
    return (
        [
            date.getFullYear(),
            padTo2Digits(date.getMonth() + 1),
            padTo2Digits(date.getDate()),
        ].join('-') +
        ' ' +
        [
            padTo2Digits(date.getHours()),
            padTo2Digits(date.getMinutes()),
            padTo2Digits(date.getSeconds()),
        ].join(':')
    );
}

// Routes
// Sends the filenames under the uploads sibling directory
app.get('/sync', (req, res) => {
    try {
        fs.appendFile('connections.txt', `${formatDate(new Date())} ${req.ip} / \n`, function (err) {
            if (err) return res.sendStatus(500)
        });

        res.json({ files: fs.readdirSync('./uploads') })
    } catch (error) {
        console.error(error);
        res.sendStatus(500)
    }
})

app.post('/upload-file', async (req, res) => {
    try {
        const files = req.files.upload
        fs.appendFile('connections.txt', `${formatDate(new Date())} ${req.ip} /upload \n`, function (err) {
            if (err) return res.sendStatus(500)
        });
        if(typeof files[Symbol.iterator] === 'function'){
            for(const file of files){
                console.log('saving: '+file.name)
                await file.mv(`./uploads/${file.name}`)
            }
        }else{
            console.log('saving: '+files.name)
            await files.mv(`./uploads/${files.name}`)
        }
        res.status(200).json({ files: fs.readdirSync('./uploads') })

    } catch (error) {
        console.error(error);
        res.sendStatus(500)
    }

})

app.delete('/delete', async (req, res) => {
    try {
        const filename = req.query.filename
        fs.appendFile('connections.txt', `${formatDate(new Date())} ${req.ip} /delete?filename=${filename} \n`, function (err) {
            if (err) return res.sendStatus(500)
        });
        await fs.promises.rm(`./uploads/${filename}`)
        res.status(200).json({ files: fs.readdirSync('./uploads') })
    } catch (error) {
        console.error(error);
        res.sendStatus(500)
    }

})

// Sends a file contents to Browser
app.get('/retrieve', (req, res) => {
    try {
        const filename = req.query.filename
        fs.appendFile('connections.txt', `${formatDate(new Date())} ${req.ip} /retrieve?filename=${filename} \n`, function (err) {
            if (err) return res.sendStatus(500)
        });
        console.log(filename)
        res.sendFile(path.join(__dirname, 'uploads', filename))

    } catch (error) {
        console.error(error);
        res.sendStatus(404)
    }
})

app.listen(9001)

Add the Frontend

For the Front-end, I have created a repo. I have added the steps in the repo but I will write them again. If you are struggling with these steps the steps in the repo are more detailed

Numero uno thing you want to do is replace every instance with your public ip ( (You can find your public ip here)[https://ifconfig.me/] )

A notable thing that I did was that in vite.config.js I added



server: {
    host: true
  }

This allows the router ( and therefore any other device in LAN ) to enter the app.

Then you want to build the app for production:
npm run build
This will generate the dist which you want to put in the server directory and add the following code to the index js file.
app.use(‘/’, express.static(‘dist’)).

Here is how it looks:



 // ...
app.use(fileupload({
    useTempFiles: true,
    tempFileDir: "/tmp/"
}))
app.use(express.json())
app.use(cors({ origin: "*", }))
// The line below serves the dist folder ( app ) when users make a request to /
app.use('/', express.static('dist'))
 // ...

Hosting your Server with port Forwarding.

These steps will differ from person to person because each of us has a different Internet Service Provider (ISP). Meaning the gateway website layout is different, I will give you the objectives and you will try to accomplish them.
1. Find your Default Gateway. On windows type ipconfig and search for Default Gateway. On UNIX systems type ip route show and you will see default via … .
2. Paste the number on the URL search bar.
3. Log in with your credentials. You will find the credentials on your router and typically the User Name is admin
4. Search for either Port Forwarding. In my case it was under NAT.
5. Add a rule, the layout will probably differ.

You are ready to fire up the server

With that, fire up the server and if you have any problems you can’t fix please contact me. Thank you very much for following along with this tutorial, now you have a file Sharing App on the web everyone you want can join, moreover you have installed trackers so if there are any uninvited people you can see them and if you want you can add an if statement checking for blacklisted people.

React.js + AI: How to combine AI with React without any Machine Learning experience.

Chris Kay — Thu, 18 May 2023 18:24:22 +0000

You are a web developer with not much Machine learning knowledge, how can I use Tensorflowjs on the Frontend you ask. There are all sorts of pretrained models for almost every case.

Thankfully Tensorflowjs and Tensorflow in general provide us with pretrained models that we can use without us making convolutional neural networks or recurrent neural networks with attention mechanisms. It is as simple as importing a library and calling a method. That’s exactly how you can use a computer vision model that can recognize 1000 objects in the Frontend without ever having heard the phrase computer vision in your career. Don’t trust me? Here you can find all Tensorflowjs pretrained models that are specialized in: audio, text, image etc. Find a niche that you would like to use in the future and let’s set up a small application using Reactjs.

Now that you have found what AI you would like to use, press the card of the AI and that will redirect you to the model’s Github page where it gives you instructions on how to import, load and use the model. Scroll down to Usage with NPM and find the npm package of the model, it will probably be “@tensorflow-models/”, I will be using the mobilenet model which the first model under vision. This section will also be guiding you on how to make prediction, so don’t close the tab! Now we are ready to code!

As usual create a react project, I will use typescript. Install the necessary dependencies:

# I will use the mobilenet model, just change the model name at the end
npm i @tensorflowjs/tfjs @tensorflow-models/mobilenet

yarn add @tensorflowjs/tfjs @tensorflow-models/mobilenet

NOTE: Although mobilenet’s Github page says it is not necessary to import the @tensorflowjs/tfjs package, when I don’t import it, it throws this error: Uncaught (in promise) Error: No backend found in registry. Even if you don’t use mobilenet beware about this error.

I am going to build a simple website that the user gives it an image and the mobilenet model predicts what that image is. Here is the Github repo for the project. Install the dependencies, run the server and start improving the code, I have made the code bad from design so you can improve it and when you are done, send me the final code.

If you are curious about AI, ML (Machine Learning) etc. I suggest you try creating your own models, at first it will be hard to create your own neural networks with all the parameters you have, however, you will find that the models big companies use are as simple as yours but with richer resources. My own advice as a beginner ML engineer that started in web dev is to first learn calculus, linear algebra then python and finally start learning ML. You don’t have to know calculus as if you graduated with a pure mathematics degree though. If you think you don’t need more maths then start learning either Tensorflow or Pytorch, then when you think you know one of them go to the other, once you learnt one it will be easier learning the other. I haven’t got further than that, I am learning Tensorflow. I learnt Pytorch and the theory behind from Daniel Bourke, I youtuber that has 40h+ of content for beginners. Also find a project that you are legitimately interested to build. Anyway, contact me whenever you want to talk and maybe cooperate and make a project together.

I built the same Web Server in 5 languages. Here's what I found…

Chris Kay — Tue, 16 May 2023 16:26:55 +0000

I created the same File Share Server in Rust ( Actix Web ), C++ ( Crow ) , Python ( Django ) and Nodejs ( Express ). The results will probably surprise you. I should note that I come from a Nodejs ( Express ) environment and I am going to be biased to Express-like frameworks.

The languages and frameworks

1. Rust ( Actix Web )

Rust is a low-level all purposed language, it is blazingly fast with a TON of features, so many features actually that only ~10% of coders know rust. I chose Actix Web for this although I could have chosen Rocket.rs or Warp but from what I read there is no big difference between all of them. Actix follows is inspired by the Expressjs API.

2. C++ ( Crow )

C++ a 45 year old low-level object-oriented programming language that lurks in almost every electronic device today yet only 22.55% of programmers know it. There weren't many options to choose from for building this project, in fact I had 2 options: Design a library myself or use Crow. I chose Crow however one tricky thing is that are 2 versions of crow the deprecated and the maintained one, be careful to install the maintained one. Crow is actually a micro web framework, it is just a single file you import. I also used the nlohmann json library for json encoding as it came with more features.

3. Python ( Django )

We all know the snake. My first language and probably yours, almost 50% of coders know python and it is known to be an easy-to-learn high-level and slow language. I could have used flask as an alternative but Django came first to my mind. Django offers more features than flask, technically flask is a micro web framework meaning lightweight.

4. JavaScript ( Express.js )

JavaScript is the language I am most comfortable with, it is like a kid of Java and Java is like the kid of C++ so JavaScript is technically a grandson of C++. JavaScript mostly powers the web and native application. It is a high-level object-oriented language with ~65% of coders of interviewed programmers knowing it and making it the most popular language as of 2022 and probably 2023. In my opinion it is so popular because it is the only language a website can understand and if you first learn to code the Frontend with Javascript odds are you will stick with it for the Backend also.

The Results

The darker color are additional lines that I thought to not count as used lines, why? For C++ the additional lines were because Crow didn't natively support "proper" file upload, so I created my own mini/micro library which is probably very slow and bad to handle the uploading. It mimics the express-fileupload library with similar functions. It is only a single 58-line header file called crow-fileupload.h . You can find the header file in the Github repo with all the other files of the servers at the end.
For python you need to do some configuration to allow static files etc. I estimated it wouldn't take more than 10 lines to add the proper config.
So the results go as following:

Python Django: 35 + 10 = 45 lines
Nodejs Express: 53 lines
Rust Actix Web: 89 lines
C++ Crow: 124 + 58 = 182 lines!

This is exciting, contradictory and logical… let me explain what I think this graph is all about. First of all Actix Web, Crow and Django perform the same so the average time range for express should have been between 40–50ms but that's not the case at all. One mistake people may make is that this graph represents how fast a language is, however, this graph represents how fast a framework is. Specifically at serving static files, not making any server side processing. Rust and C++ are faster than JS and especially Python but frameworks do not follow that pattern, they follow the pattern of popularity most times. If you need to do some crazy big computations or handle huge traffic then you may consider picking a faster language.
Here are the results:

My sources

Statistics source for most used/popular programming languages in 2022: https://www.statista.com/statistics/793628/worldwide-developer-survey-most-used-languages/ .
Github repo for this project: https://github.com/yaKsirhC/5-lang-server/tree/master