Machine Learning Manager System Using FastAPI & Rest

This is a manager system using rest to build create, train and predict models stored in a relational database.

How To Run

First make sure you have poetry installed. Then:

$ poetry run python app/app.py

Or with docker do:

$ docker-compose up

Available Endpoints

• POST /models: Creates a model with the specified parameters in the body. Example body:

{
  "model": "MLPClassifier",
  "params": {
    "alpha": 0.0001,
    "hidden_layer_sizes": 20
  },
  "num_features": 4,
  "num_classes": 2
}

• GET /models/<uid>: Retrieves a specific model.
• POST /models/<uid>/train: Trains partially a model that has been already created. Example body:

{
  "feature_vector": [
    1.11,
    2.22,
    3.33,
    -4.44
  ],
  "label": 1
}

• GET /models/<uid>/predict?input_vector=<base64-codified-vector>: Predics the label for the input vector specified in query params.
• GET /models: Get most trained scores per each type of model.