Train the Model

In [ ]:

    

import numpy as np
from sklearn import linear_model
from sklearn import datasets

# Load the diabetes dataset
diabetes = datasets.load_diabetes()

# ONLY USING 1 FEATURE FOR THIS EXAMPLE!
# Use only one feature
diabetes_X = diabetes.data[:, np.newaxis, 2]

# Split the data into training/testing sets
diabetes_X_train = diabetes_X[:-20]
diabetes_X_test = diabetes_X[-20:]

# Split the targets into training/testing sets
diabetes_y_train = diabetes.target[:-20]
diabetes_y_test = diabetes.target[-20:]

# Create linear regression model
model = linear_model.LinearRegression()

# Train the model using the training sets
model.fit(diabetes_X_train, diabetes_y_train)


    

In [6]:

    

class PioBundle(object):

    def __init__(self,
                 model):
        self.model = model

        
    def predict(self,
                request):

        return self.model.predict(request)


    def transform_request(self,
                          request):
        import ujson
        import numpy as np
        request_str = request.decode('utf-8')
        request_str = request_str.strip().replace('\n', ',')
        # surround the json with '[' ']' to prepare for conversion
        request_str = '[%s]' % request_str
        request_json = ujson.loads(request_str)
        request_transformed = ([json_line['feature0'] for json_line in request_json])
        return np.array(request_transformed)   

    
    def transform_response(self,
                           response):
        import ujson
        return ujson.dumps(response.tolist())


    

In [ ]:

    

import dill as pickle

pio_bundle = PioBundle(model)

pio_bundle_pkl_path = 'pio_bundle.pkl'

with open(pio_bundle_pkl_path, 'wb') as fh:
    pickle.dump(pio_bundle, fh)


    

In [ ]:

    

%%bash

ls -l pio_bundle.pkl


    

In [ ]: