Sign inSign up

davidgasquez/moser

By davidgasquez

•Updated almost 10 years ago

A simple way to productionize your Machine Learning models.

Image
0

300

davidgasquez/moser repository overview

⁠Moser: A Model Server Prototype

Moser is a simple way to serve your Machine Learning models. The goal of Moser is making super easy to setup any kind of predictive model in a production server and being able to interact with them through RESTful API calls.

⁠Features

  • Set Models
  • Set Functions

⁠Get Started

You can test it running the container (make) and heading to http://localhost:5000⁠

  1. Generate the model pkl
  2. Build the container make build
  3. Run the API with make
  4. Make some requests to http://localhost:5000⁠:
    • Add a model to the API with a PUT call to /api/models/your_model_name
    • Use the previous model calling /api/models/your_model_name/predict with a POST request providing a single JSON with the feature names and values.
⁠Example

Generate the pkl file using joblib:

import pandas as pd
import numpy as np
from sklearn.datasets import load_iris
from sklearn.ensemble import RandomForestClassifier
from sklearn.externals import joblib

iris = load_iris()
data = pd.DataFrame(data=np.c_[iris['data'], iris['target']],
                    columns=iris['feature_names'] + ['target'])

clf = RandomForestClassifier()
clf.fit(data.drop('target', axis=1), data['target'])

joblib.dump(clf, 'model.pkl', compress=9)

Once we have the file we need to run the server with make and make some API requests:

import requests

base_url = 'http://localhost:5000'

filename = 'model.pkl'

# Set the model
with open(filename, 'rb') as f:
    model = f.read()
    r = requests.put(base_url + '/api/models/iris', data=model)
    print(r)

# Make predictions from JSON
data = {
    "features": ["sepal_length", "sepal_width", "petal_length", "petal_width"],
    "values": [
        [1, 4, 1, 1],
        [2, 0, 6, 1],
        [1, 4, 8, 1]
    ]
}
r = requests.post(base_url + '/api/models/iris/predict', json=data)
print(r, r.json())

Tag summary

Content type

Image

Digest

Size

308.6 MB

Last updated

almost 10 years ago

docker pull davidgasquez/moser