DurianDan / demo-mlops-workflow

workflow for a MLOps system, attending in "MLOps Marathon 2023"

1. Project Structure

• api: receive and response to MLOpsVN's requests
  • main.py: endpoint of the API, execute uvicorn main:app to run API
  • problem
    • api_models: Define model classes, to interact with SQL tables.
      • base.py: define Base, a base class for all models to inherit
      • factory: create Specific model based on phase_id and prob_id
  • schema: define request/response body

Data Scientist-ing

• To serve a model, put the model in this folder: flow/api/predict/MLmodels
  • E.g.: to save model sample.model.file in phase 2, to solve problem 3

    flow/api/predict/ml_models/phase2/prob3/