GraphicalEditorForNN

Links

Figma

Ugile project (authorization required)

Schedule table

Retrospectives

Artifacts

About

Our project is a visual editor for neural networks designed to simplify the creation and customization of deep learning models. In a browser-accessible interface, users can intuitively assemble, configure, and optimize neural networks by dragging layers and adjusting parameters in real-time.

A key feature of the project is the ability to train models directly from the editor. The C++ server ensures high performance, allowing users to experiment with architectures and parameters. The trained neural networks generated are ready to predict outcomes for new data

Features of our project:

• Intuitive Interface: User-friendly experience thanks to an intuitively designed interface.
• Progress Saving: Capability to save the current work progress for user convenience.
• Light/Dark Themes: Personalization of the user experience.
• Custom Framework for Neural Network Training: Utilization of our own framework with automatic gradient computation for efficient neural network training.
• On-the-Spot Training and Prediction: Ability to conduct training sessions and obtain predictions directly within the neural network editor.
• Export Capabilities: The option to export created models.
• Neural Network Integrity Check: Built-in validation of the correctness of the neural network.

Build options

Important: If CI tests don't pass, but work locally, make sure your branch is updated (rebase on the current main)

Important: All Makefiles require node to run.
On MacOS it is just brew install node.
On linux run curl -qL https://www.npmjs.com/install.sh | sudo sh and you should be good

There is one Makefile at the project root and a Makefile for each folder (client, py_server, server)

For each Makefile you can run

make help

It will give a list of all available targets and options for the current Makefile

Common Options (make <something> option1=value1 option2=value2 ...):

• V: verbose level, if you want make to print what it's doing use V=2
• CONFIG_PATH: path to the config.json, default is the right one

Also, if you need some other options, remember that you can override any variable used in the Makefile using an option with that name.

Run Client

Client has only one target. You can also provide options listed in the make help for example:

cd client
make port=2002 V=2

Runs server on the port 2002, or just use make to run it on the default port

Important: npx serve is used to host the client code (index.html) on the port specified in config.json
Python server is configured to only accept requests from this port
If you run your index.html as a file and will send requests to python server, CORS won't allow you to do it
If you want to make it simple and allow Python server to accept anything, uncomment line at __init__.py:41

Run Python Server

Make sure you have python version at least 3.10

Important: when you first install this new build system, delete old virtual environment in py_server folder with rm -r .venv. Then, if your default python command uses python version lower than 3.10, specify the python interpreter explicitly: make python=python3.10 for example. You only need to specify it explicitly once, after that make will use python from the created virtual environment.
make clean will not delete .venv folder, so if you created it with a wrong python version, delete it and then run make python=... again with a correct version of python

If you see error like this on linux, when trying to create a .venv:

Error: Command '['.../.venv/bin/python3.11', '-m', 'ensurepip', '--upgrade', '--default-pip']' returned non-zero exit status 1.

Try sudo apt install <your python path>-venv

Then a simple make in the py_server directory should work
It will build you a virtual environment in the folder py_server/.venv. Then it will install all dependencies in it via pip, including our own project mlcraft.
After that, it launches mlcraft using flask at specified port (see make help for all available options here).

cd py_server
make 

Other targets available:

• make format: formats your code using black formatter, use before commit, CI won't let you pass with unformatted code
• make test: launch mypy typechecking and all tests from tests directory using pytest
• make clean: deletes current instance folder of the app (I don't think you'll need to use it ever)

Run C++ Server

Important: If you want to build a c++ server, you need to install Boost.
Then, in the config.json file add the path to the boost root (folder with include and lib inside).
For example: "BOOST_ROOT": "/usr/local/Cellar/boost/1.81.0_1"

Also you need cpprest: on MacOS: brew install cpprestsdk
on Linux: sudo apt-get install libcpprest-dev
After that you should be able to build everything just fine...

There are 3(4) main targets available to build:

• make: runs the core/main.cpp file (use it to test your Tensor and so on...)
• make serve: runs the server (api/server.cpp)
• make test: runs all the tests in the tests directory (test file for X.cpp must be named XTests.cpp)
• make test.X: runs a single test for X

Examples:

cd server
make serve port=4000
make test.Blob V=2 O=2

To change the compiler flags you can edit some Makefile lines about the CXXFLAGS variable

Run from root folder

You can build and test things from the root folder as well, but I don't think it is useful. It's done only to conveniently run the CI
Use make help to see available targets there

Technology stack

Languages:

• C++
• Python
• Javascript

Server++:

• Network: Crow
• Build: Makefile

Server:

• Flask
• SQLAlchemy

---

Architecture

Overview

Server++

Data