Flask server running a XOR Keras model.
This very basic Keras model learns the XOR operation. Run the model using python train.py.
from keras.models import Sequential
from keras.layers.core import Dense
from keras.optimizers import SGD
import numpy as np
X = np.array([[0,0],[0,1],[1,0],[1,1]])
Y = np.array([[0],[1],[1],[0]])
model = Sequential()
model.add(Dense(8, input_dim=2, activation='tanh'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer=SGD(lr=0.1))
model.fit(X, Y, batch_size=1, nb_epoch=1000)
model.save('xor_model')Run the server using python server.py.
from flask import Flask, request
from keras.models import load_model
import tensorflow as tf
import numpy as np
import flask
app = Flask(__name__)
graph = tf.get_default_graph()
model = load_model('xor_model')
@app.route('/predict')
def predict():
a = request.args['a']
b = request.args['b']
with graph.as_default():
result = model.predict(np.array([[a,b]]))[0].tolist()
data = {'result': result}
return flask.jsonify(data)
app.run(host='0.0.0.0', debug=False)You can make a GET request using your browser :
http://ip_address:5000/predict?a=0&b=1
