Deploying a Tensorflow Model

Setup

Tensorflow models are slightly different from "regular" Python models. Tensorflow requires a custom runtime environment. You'll need to configure your ScienceOps cluster to use the yhat/scienceops-tensorflow:0.10.0 (or newer) base image.

Example Model

In this example we'll be using the linear regression form the tensorflow pre-made examples.

Import the tensorflow module and begin training:

import tensorflow as tf
import numpy as np

# Parameters
learning_rate = 0.01
training_epochs = 10
display_step = 50

# Training Data
train_X = np.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,
                         7.042,10.791,5.313,7.997,5.654,9.27,3.1])
train_Y = np.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,
                         2.827,3.465,1.65,2.904,2.42,2.94,1.3])
n_samples = train_X.shape[0]

# tf Graph Input
X = tf.placeholder("float")
Y = tf.placeholder("float")

# Set model weights
W = tf.Variable(np.random.randn())
b = tf.Variable(np.random.randn())

# Construct a linear model
pred = tf.add(tf.mul(X, W), b)

# Mean squared error
cost = tf.reduce_sum(tf.pow(pred-Y, 2))/(2*n_samples)
# Gradient descent
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

# Initializing the variables
init = tf.initialize_all_variables()

# Launch the graph
sess.run(init)

# Fit all training data
for epoch in range(training_epochs):
    for (x, y) in zip(train_X, train_Y):
        sess.run(optimizer, feed_dict={X: x, Y: y})

    # Display logs per epoch step
    if (epoch+1) % display_step == 0:
        c = sess.run(cost, feed_dict={X: train_X, Y:train_Y})
        print("Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(c), \
            "W=", sess.run(W), "b=", sess.run(b))

print("Optimization Finished!")
training_cost = sess.run(cost, feed_dict={X: train_X, Y: train_Y})
print("Training cost=", training_cost, "W=", sess.run(W), "b=", sess.run(b), '\n')

Deployment

To deploy your model, you'll use the same methodology as a regular Python model with the exception of the deploy_tensorflow command.

Tensorflow works differently than scikit-learn or numpy, and as a result you need to explicitly define the variables you'll be using from your tensorflow model. To do so, define a setup_tf function in your YhatModel class. setup_tf should contain the definitions of any tensorflow related variables you'll be using to make a prediction.

test_X = np.asarray([6.83, 4.668, 8.9, 7.91, 5.7, 8.7, 3.1, 2.1])
print(sess.run(W) * test_X + sess.run(b))

from yhat import YhatModel, Yhat

class TfModel(YhatModel):

    REQUIREMENTS=['tensorflow==0.10.0','numpy==1.11.1']
    def setup_tf(self):
        import tensorflow as tf
        import numpy as np
        # tf Graph Input
        X = tf.placeholder("float")
        Y = tf.placeholder("float")

        # Set model weights
        W = tf.Variable(np.random.randn())
        b = tf.Variable(np.random.randn())

    def execute(self, data):
        y_pred = sess.run(W) * np.array(data['x']) + sess.run(b)
        return {"y": y_pred.tolist()}

yh = Yhat("USER", "API_KEY", "URL")
yh.deploy_tensorflow("tensorflowExample", TfModel, globals(), sess)

Pass your tensorflow session variable (should be sess) to the deploy_tensorflow function in the yhat client. This will allow yhat to serialize your model and upload it to ScienceOps.

results matching ""

    No results matching ""