Frosendroska/nnbuilder

Web application allowing to construct neural network and train datasets without coding

nnbuilder

About a project

WEB PLATFORM FOR RESEARCHERS TO HELP THEM MAKE A NEURAL NETWORK USING A PAINLESS VISUAL CONSTRUCTOR AND THEN UPLOAD A DATASET TO GET A TRAINED MODEL

Run in prod

1. Add DB parameters to .env

2. gw build -x test

3. java -jar nnbuilder/build/libs/nnbuilder-0.0.1-SNAPSHOT.jar

Development setup

0. Install JDK 17, Gradle, Node, Docker; alias ./gradlew to gw
1. Clone the repository
2. Write a .env config file (or copy from .dev.env example)
3. Run gw idea assemble to prepare the Intellij project and build intermediate files
4. Run gw dockerComposeUp to enable proxy
5. Open nnbuilder.ipr file in Intellij (on macOS run open *.ipr)

Install database on MAC

• brew install postgres installing postgresql
• pg_ctl -D /usr/local/var/postgres start start a database
• psql postgres go inside database
• postgres=# create database nnbuilder; make a database
• postgres=# create user nnbuilder with encrypted password 'nnbuilder'; set a password
• postgres=# grant all privileges on database nnbuilder to nnbuilder; grant all privileges
• Then go to IDEA and tab on database->+ and add new database with created password and username

Install database on Linux

• install postgresql
• sudo -u postgres psql start postgresql
• postgres=# create database nnbuilder; make a database
• postgres=# create user nnbuilder with encrypted password 'nnbuilder'; set a password
• postgres=# grant all privileges on database nnbuilder to nnbuilder; grant all privileges
• Then go to IDEA and tab on database->+ and add new database with created password and username

Helpful commands:

• gw build performs a full build
• gw nnbuild-api:assemble re-generates grpc code from protobuf definitions
• gw spotlessApply applies code style fixes
• gw dockerComposeDown to disable proxy