Installation

Requirements

• Linux or macOS with Python >=3.6

• PyTorch and torchvision

• Node.js

Build - Inital Step

# get the latest version
git clone https://github.com/patilli/vqa_benchmarking.git

Python Backend

The backend is written in Python for PyTorch. It provides datasets and model adapters to be integrated into any PyTorch repository. The data gets stored in sqlite3 databases with a tornado web server.

Build

We recommend to create and activate a new python virtual environment for the following instructions.

# change directory
cd vqa_benchmarking/vqa_benchmarking_backend

# install required libraries
pip install -r requirements.txt

# install (consider adding -e)
pip install .

Web Interface

The web application is written with vue.js. It requires a javascript package manager like npm to be installed on your machine.

Build

# change directory
cd vqa_benchmarking/vqa_benchmarking

# install dependencies
npm install

# serve with hot reload at localhost:8080
npm run dev

# build for production with minification
npm run build

# build for production and view the bundle analyzer report
npm run build --report