FOG Computing Project

Welcome!

This repo was made as part of FOG Computing course lead by prof. dr. Vlado Stankovski.

Architecture

Tracking/managment of our experiments was done using Weights & Biases.

We are packaging up the prediction system and deploying it as a Docker container on AWS Lambda.

Our prediction system is wrapped in a frontend written in Python using Gradio.

Colab Demo

Local installation

conda env create -f environment.yml
wandb login <WANDB_API_KEY>
cd src

Local deployment of the frontend and backend

PYTHONPATH="$PWD" python training/stage_model.py --fetch
PYTHONPATH="$PWD" python app_gradio/app.py

Local deployment of the frontend and AWS backend

PYTHONPATH="$PWD" python app_gradio/app.py \
  --model_url https://3akxma777p53w57mmdika3sflu0fvazm.lambda-url.us-west-1.on.aws/

https://github.com/the-full-stack/fsdl-text-recognizer-2022-labs

Acknowledgement

Our code is largely based on the following open-source projects: FSDL-Text-Recognizer.