BREVIS-News 2.0

Original unfinished project Brevis-News

• BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension: https://arxiv.org/abs/1910.13461

Current Status:

Completed Features:

• Created custom serving model based on BART, pre-trained model from huggingface and fine-tuned on inshorts data
• Created Django backend for:
  • Article scraping
  • Google-News API trending news feed
• Single line docker-compose deployment
• Designing front-end

Pending Work:

• Overhaul front-end
• Fine-Tune model on more dataset
• Heroku Deployment
• Trending news auto-summarization

Usage

With docker-compose

Just run docker compose up and the everything should just work.

• Does not work on windows due to WSL not supporting host mode on the port

docker-compose up

It will take a long time for the first run

Without docker-compose

BART Serving:

Step 1: Clone this repo

git clone https://github.com/vinayak19th/MDD_Project

Step 2: pull the docker container

For information on installing docker, check here

docker pull vinayak1998th/bart_serve:1.0

Step 3: Launch the container

CPU Runtime

docker run -d -p 8501:8501 -p 8500:8500 --name bart vinayak1998th/bart_serve:cpu

If you have an NVIDA CUDA supported GPU, you can run the server for GPU runtime

GPU Runtime

docker run --runtime=nvidia -d -p 8501:8501 -p 8500:8500 --name bart vinayak1998th/bart_serve:gpu

Step 4 : Test

The code for testing after the server is running is in the Serving_Test notebook

Contributing to this Project

Read Contributing.md file in the repository to learn more about making pull requests, and contributing in general.

Project Maintainers

The project is currently being maintained by :

1. Vinayak Sharma