mlops-trainee-udacity-course-4

project for udacity mlops lecture4, build a simple Decision Tree model on census data.

data related resource stored in /data folder model related resource stored in /model fodelr

using dvc local_remote mode.

Install requirements

you can use poetry with conda. install conda environment first.

conda create -n mlops python=3.10
conda activate mlops
pip install poetry-conda

then

poetry install

or use pip

pip install -r requirements

[exception]: add local remote

mkdir ../local_dvc_remote
dvc remote add -d local_remote ../local_dvc_remote

Machine Learning running

Run simple ml pipeline by using mlflow+hydra

mlflow run . -P steps=data_cleaning,training

or run step by step

mlflow run . -P steps=data_cleaning

Test

pytest test_main.py

and

pytest src/training/test_model.py

Project Rubic

git

• Set up git with GitHub Actions.

model buliding

• Create a machine learning model, check src/training folder
• Write unit tests, check test under src/training
• Write a function that computes model metrics on slices of the data. check funciton src/training/model.py plot_model_disparity_on_fpr
• Write a model card. check modelcoard under /model folder

API Creation

• Create a REST API. check main.py
• Create tests for an API. check test_main.py

API Deployment

• Deploy an app to a Cloud Application Platform. check png under /screenshots folder
• Query live API. check png under /screenshots folder. run request_sample.py