FastAPI for student grade prediction

Данг Куинь Ньы (Нина), tg: @quynhu_d

Dataset taken from kaggle.

Objective: predict student grade based (regression).

HW1

Check branch hw1_checkpoint.

To start:

Create virtual environment, install required packages:

python -m venv fastml_env
source fastml_env/bin/activate
pip install -r requirements.txt

Run API:

cd src
uvicorn app:app --reload --host 0.0.0.0 --port 8008

or

cd src
python app.py

HW2

Check branch hw2_checkpoint. Added minio functionality, tried dvc.

App docker image: quynhud/grade-prediction-app:latest.

To run application:

docker-compose up -d

Minio s3 at http://localhost:9001/.

HW3

Unit tests in src/test_app.py

pytest src/test_app.py

Loading and saving functions for s3 are mocked.

CI with Github Actions: .github/workflows/ci.yml:

• Builds and pushes Docker images to Docker Hub only upon merge requests
• Flake8 linter (allows lines up to 100 characters)

Documentation

API with Swagger Documentation can be accessed via http://localhost:8008/docs#/.

Functionality:

1. list_models List available models.
2. list_saved_models: List saved trained models.
3. list_saved_datasets: List saved datasets.
4. save_data: Save datasets.
5. train: train regression model (see documentation via swagger)
6. predict: predict using fitted model (see documentation via swagger)
7. retrain: retrain previously fitted model (see documentation via swagger)
8. delete_model: delete existing model checkpoint
9. delete_data: delete existing data

Example:

Pass data/train_data.json as data and examples/model_configs/svr_config.json as params to train upon query.

All parameters must be filled in, see LinearRegressionConfig/RandomForestRegressorConfig/DecisionTreeRegressorConfig/SVRConfig specification in Schemas and/or example values in examples/model_configs.