MLOps Pipeline for Cat vs Dog Dataset

Currently, this repo is following Stage 1. Read here.

app/                 - application deployment
└── app.py           - run application

data/                - data pipeline
├── params           - Parameter
    └── param.yaml   - parameter file
├── log              - log folder
├── figures          - figures folder
├── notebooks        - jupyter notebooks
├── eda.py           - exploratory data analysis
├── preprocess.py    - preprocessing data
└── validate.py      - validating data

storage/
├── artifact         - model 
├── data-lake        - raw data
└── data-warehouse   - preprocessed data 

Environment setup

Python version 3.8.5 is installed.

python -m venv pipeline-env
pip install -r requirements.txt

N.B: If you are in wsl, make sure run pip install wheel if needed.

GCloud

Install GCloud SDK from here.

gcloud auth login
gcloud auth application-default login

DVC

dvc remote add -d myremote gs://ml-pipeline-demo-storage
dvc push <file_name.dvc> -r myremote
dvc pull <file_name.dvc> -r myremote

MLflow

Will run the mlflow server in a central server. If you want to run mlflow locally, just run mlflow ui from root folder.

Airflow

Use wsl or linux as there is no support for windows. Read setup in here.