Note: based on this tutorial.
Setting up an environment
python3.11 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
Run a local tracking server
mlflow server --host 127.0.0.1 --port 8081
Open http://127.0.0.1:8081 in a browser
In another terminal window
source venv/bin/activate
export MLFLOW_TRACKING_URI=http://127.0.0.1:8081
cd 1_localhost
python local_deployment_demo.py
In the UI choose a model based on a metric
And deploy the selected model, setting a name
And deploy the model, spefifying the
mlflow models serve -m models:/sample_model/1 --port 5001
Note 1: you might need to handle the issues like in this link, running the commands below.
curl https://pyenv.run | bash
python -m pip install virtualenv
PATH="$HOME/.pyenv/bin:$PATH"
Note 2: If the deployment command above does not work, you can deploy the model also by specifying the model path like -m <path-of-a-model> instead of models:/<model-name>/<model-version>
In another terminal window, curl the endpoint just made
curl -d '{"dataframe_split": {
"columns": ["fixed acidity","volatile acidity","citric acid","residual sugar","chlorides","free sulfur dioxide","total sulfur dioxide","density","pH","sulphates","alcohol"],
"data": [[7,0.27,0.36,20.7,0.045,45,170,1.001,3,0.45,8.8]]}}' \
-H 'Content-Type: application/json' -X POST localhost:5001/invocations
If you get a output like below, it is a success
{"predictions": [[5.719637393951416]]}%