Page tree
Skip to end of metadata
Go to start of metadata

Data preparations

  • MIS (Management Information Services) team supports with data related requests and data set creations
  • Data warehouse is the central data repository for data, which should be the go-to-place for data
  • Data sets should be reproducible, ideally as SQL or API calls

Model development

  • Use virtual environment to manage dependencies / libraries
  • Use requirements.txt to store list of dependencies
  • "pip freeze > requirements.txt" to create requirements.txt file with needed libraries
  • "pip install -r requirements.txt" to install the libraries in your project environment
  • Version control should always be used to manage source code, including trained models, scripts, training data and project configurations. Aalto uses version.aalto.fi Git version control system
  • Jupyter Lab / Notebook is used for model development during experimentation phase
  • SW IDE / editor (e.g., Visual Studio Code) for finalizing and testing the models as python .py files
  • Azure ML Studio can be used for creating simple tasks using drag-and-drop functionality

Model deployment

Models should be deployed and made available as micro-services published as APIs.

  • No labels