sample-model-server

Install the Requirement

Suggest to run this in a virtual environment then install the requirements.

pip install -r requirements.txt

To Run a Prediction via a Web Browser

1. Fire up a terminal in this directory and run this command

   export FLASK_APP=app.py && flask run 1.1 To run in https export FLASK_APP=app.py && flask run --cert=cert.pem --key=key.pem

2. Then navigate to localhost:5000/predict

3. Upload the data file in ./data (Suggest to try pickle_scikit_lr_compas_xtest_pkl.sav)

4. Wait for results

To Start an API Server to Test Run Models With Inputs From an Algorithm

1. cd to api-model-server directory and run:

   source venv/bin/activate

2. To run a regression model server, run:

   bash regression_run_http.sh to start a HTTP server on port 5000

   and

   bash regression_run_https.sh to start a HTTPS server on port 5000

3. To run a classification model server, run:

   bash classification_run_http.sh to start a HTTP server on port 5001

   and

   bash classification_run_https.sh to start a HTTPS server on port 5001

To Run a Prediction via API

1. Fire up a terminal in this directory and run this command

   flask app

2. Run this curl command (replace the file path accordingly)

   curl -X POST -F file=@"./data/pickle_scikit_lr_compas_xtest_pkl.sav" http://localhost:5000/predict

3. Or run this curl command (replace the value accordingly)

   This will return 0 curl -X POST http://localhost:5000/predict_one -H "Content-Type: application/x-www-form-urlencoded" -d "count=0&gender=0&race=5&charge=1&age=2"

   This will return 1 curl -X POST http://localhost:5000/predict_one -H "Content-Type: application/x-www-form-urlencoded" -d "count=10&gender=1&race=5&charge=1&age=2"

4. Or run this curl command (replace the value accordingly) for mulitple dataset

   curl -X POST http://localhost:5000/predict_all -H 'Content-Type: application/json' -d '{"input": "[[0, 1, 0, 0, 0], [0, 1, 0, 0, 0], [0, 1, 3, 11, 1], [0, 1, 2, 0, 0], [0, 1, 5, 2, 0], [1, 0, 0, 1, 0], [0, 1, 5, 0, 0], [0, 1, 0, 1, 1], [0, 1, 0, 9, 0], [1, 1, 0, 1, 0]]"}'

Others

1. Note that this is not the only way to write a model server

   • The response might differ from server to server.
   • The way to upload the input data could differ too.