krishnakalyan3 / pets-app-demo

Introduction

Lightning App is composed of Lightning Work and Lightning Flow. Diagram to illustreate this below. Also please feel free to check out a related lightning-hpo component.

Creating your environment

This is a demo lightning app that gradually shows us how to build lightning applications step by step. Make sure that you execute the commands below.

conda create -n wnb python=3.9 
conda activate wnb
pip install lightning
git clone https://github.com/krishnakalyan3/pets-app-demo
cd pets-app-demo
pip install -r requirements.txt

Execution on Cloud

This is a demo lightning app that gradually shows us how to build a lightning application step by step. Serial execution is adviced as complexity varies. After running the training-app please copy the sweep ID (which begins like wandb agent krishnakalyan/pets/3xhnzyb4) and replace it in the sweep application (02_sweep_app.py). You can modify the sweep application have more GPUs for your sweep experiment as well as number of sweeps.

The last example 03_serve_app.py shows us a way to serve this application using gradio.

# Create Secrets (TODO)
WANDB_API_KEY
KAGGLE_USERNAME
KAGGLE_KEY

# Jupyter Application
lightning run app 01_jupyter_app.py --env =$WANDB_API_KEY \
--env KAGGLE_USERNAME=$KAGGLE_USERNAME --env KAGGLE_KEY=$KAGGLE_KEY \
--cloud --open-ui false --name training-app

# Sweep Application
lightning run app 02_sweep_app.py --env WANDB_API_KEY=$WANDB_API_KEY \
--env KAGGLE_USERNAME=$KAGGLE_USERNAME --env KAGGLE_KEY=$KAGGLE_KEY \
--cloud --open-ui false --name sweep-app

# Streamlit Application
python -m lightning run app 03_serve_app.py --open-ui false --name streamlit-app \
--env WANDB_API_KEY=$WANDB_API_KEY  \
--cloud --open-ui false --name wnb-serve

Known Issues

• Please sign in to access your W&B report.