IGI.MachineLearning

Note: This is intended for experimentation / proof of concept work, not for production use.

Experiment with machine learning models and expose machine learning models via a web service.

---

Local Web Service via packaged exe

Originally implemented with IGI.DataServices to plug file transformation into p:IGI+ / Transform, but the same principle of wrapping a python env, libraries and project into an executable that starts a local web service could be used for the machine learning project if required (if clients are concerned about sending data).

Build local runner exe

Use auto_py_to_exe or if using pyinstaller from cmd line run the command below (under creating the exe sub heading). Note: the local python evnironment used when creating the exe will determine which python version and which libraries are bundled with this code into the exe.

This should be run from a virtual environment to ensure that only libraries required for this project are included.

Setting up a virtual environment

Install Virtual env if you don't already have it:

pip install virualenv

Create an environment e.g. from the solution root, to create the virtual env in a .env subsir (ignored by github):

python -m venv .venv

Activate the virual env:

.\.venv\Scripts\activate (you should see an indication in your shell that you are now in the venv)

Install all libraries required for the project:

pip install -r .\requirements.txt

Install pyinstaller library (used to make the exe) - install in this venv even if you have it in your main environment (otherwise you main env will be used to create the exe):

pip install pyinstaller to confirm that you will use the version for ths env enter the powershell comand Get-Command pyinstaller and you will see the source path used.

Install a prod http server (waitress) - similar to gunicorn, but works on windows. If this step is skipped it will revert to the builtin flask wsgi server - which works, but is intended for dev/debugging not production. This is not listed in the requirements.txt as it is not useful for all use cases of this codebase.

pip install waitress

To deactivate the virtual env (after running the command to create the exe below) use:

deactivate

Creating the exe

pyinstaller --noconfirm --onefile --console --icon "D:/code/projects/IGI.MachineLearning/pigipy_128.ico" --name "pigipy"  "D:/code/projects/IGI.DataServices/local_runner.py"

change --console to --windowed if you don't want the app to bring up a console window (i.e. to hide).

Running this command writes the output exe to ./dist by default.

---

Example vscode config (for local run)

launch config for ${workspaceRoot}/.vscode/launch.json:

{
    "version": "0.2.0",
    "configurations": [
        {
            "name": "Flask debug",
            "type": "python",
            "request": "launch",
            "stopOnEntry": false,
            // need to add app.run(port=5000) to file for this to work - 
            // allows debugging / break points etc in vscode
            "program": "${workspaceRoot}/app.py", 
            "env": {
                "FLASK_APP": "${workspaceRoot}/app.py",
            },
            "args": [
                "local debug",
            ],
            "envFile": "${workspaceFolder}/.env",
            "redirectOutput": true,
            "jinja": true
        },
    ]
}