/rapids

http://rapids.ai

Primary LanguageJupyter Notebook

Rapids End to End ML Demo

Contents

  1. Motivate rapids [ show coverage of modern data science tools ]

  2. Generate a synthetic dataset

  • 2.1 - Split into train and test set
  • 2.2 - Visualize sub-datasets
  1. ETL
  • 3.1 - Load data [ csv read ]
  • 3.2 - Transform data [ standard scaler ]
  1. Model Building
  • 4.1 - Train CPU and GPU XGBoost classifier models
  • 4.2 - Use trained models for inference
  • 4.3 - Compare accuracy
  • 4.4 - Visualize sample boosted trees & model predictions
  1. Extensions
  • 5.1 - Create an ensemble with a clustering model [ DBScan ]
  • 5.2 - Export data to DeepLearning Framework [ PyTorch ]

Install & Run Demo

Video Walkthrough Link

1 -- clone repository

git clone https://github.com/miroenev/rapids && cd rapids

2 -- build container [ takes 5-10 minutes ]

sudo docker build -t rapids-demo:latest .

3 -- launch/run the container [ auto starts jupyter notebook ]

sudo docker run --runtime=nvidia -it --rm -p 8888:8888 -p 8787:8787 rapids-demo:latest

4 -- connect to notebook

i) navigate browser to IP of machine running container at port 8888
    e.g., http://127.0.0.1:8888

ii) in the rapids folder launch the notebook titled 
    rapids_ml_workflow_demo.ipynb