By using Vertex AI by GCP one can deploy models easily that are trained by frameworks like scikit-learn, pytorch, tensorflow and xgboost.
Dataset taken from Kaggle: https://www.kaggle.com/datasets/uciml/pima-indians-diabetes-database
The GradientBoost model in pickle format is available in the repo so you can just clone the repo to get started.
git clone https://github.com/ChiragChauhan4579/Diabetes-Predicor-GCP
To install required libraries run this command
pip install -r requirements.txt
Lets start by storing the models pickled file on cloud storage. To do this search for Cloud Storage on GCP console and then create a bucket there with your required configurations. After createing the bucket you can see various upload options so upload your pkl file through that.
- Search
Vertex AIand go to the model registry through the left hand pane in the there. Click onImportbutton to start registering a new model by giving it a name. - In model settings section choose the
framework, its version (here the version is 0.20) and pickled model location. - If you want explainability you can choose it or skip.
- Finally click the
import button. Model import will take some time. - Create the endpoint of that model by going to endpoint section in left pane.
Note: Your model name should be strictly model.pkl
- Install the cli using this link
- Create the credential file as mentioned in instructions using
gcloud auth application-default login
Run the application after changing the file as
- Give the
endpoint deployed regionin predict_class argument api_endpoint. - Input your details when calling predict_class function by going to sample request -> python in created endpoint of vertex ai.
predict_class( project="ENTER YOUR DETAILS", endpoint_id="ENTER YOUR DETAILS", location="ENTER YOUR DETAILS", instances=[preg,gluc,bp,sth,insu,bmi, dpf,age])
streamlit run app.py
Output:
