In [1]:

    

import time
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf


    

In [2]:

    

mnist = input_data.read_data_sets('', one_hot = True)


    

    




Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.
Extracting train-images-idx3-ubyte.gz
Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.
Extracting train-labels-idx1-ubyte.gz
Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.
Extracting t10k-images-idx3-ubyte.gz
Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.
Extracting t10k-labels-idx1-ubyte.gz

In [3]:

    

class Model:
    
    def __init__(self, learning_rate, y_shape):
        self.X = tf.placeholder(tf.float32, (None, 28, 28, 1))
        self.Y = tf.placeholder(tf.float32, (None, y_shape))
        
        def convolutionize(x, conv_w, h = 1):
            return tf.nn.conv2d(input = x, filter = conv_w, strides = [1, h, h, 1], padding = 'SAME')

        def pooling(wx):
            return tf.nn.max_pool(wx, ksize = [1, 2, 2, 1], strides = [1, 2, 2, 1], padding = 'SAME')
        
        w1 = tf.Variable(tf.random_normal([3, 3, 1, 16], stddev = 0.5))
        b1 = tf.Variable(tf.zeros(shape = [16]))
        w2 = tf.Variable(tf.random_normal([3, 3, 16, 8], stddev = 0.5))
        b2 = tf.Variable(tf.zeros(shape = [8]))
        w3 = tf.Variable(tf.random_normal([3, 3, 8, 8], stddev = 0.5))
        b3 = tf.Variable(tf.zeros(shape = [8]))
        w4 = tf.Variable(tf.random_normal([128, y_shape], stddev = 0.5))
        b4 = tf.Variable(tf.zeros(shape = [y_shape]))

        conv1 = pooling(tf.nn.sigmoid(convolutionize(self.X, w1) + b1))
        conv2 = pooling(tf.nn.sigmoid(convolutionize(conv1, w2) + b2))
        conv3 = pooling(tf.nn.sigmoid(convolutionize(conv2, w3) + b3))
        conv3 = tf.reshape(conv3, [-1, 128])
        self.logits = tf.matmul(conv3, w4) + b4
        
        self.cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = self.logits, labels = self.Y))
        self.optimizer = tf.train.AdamOptimizer(learning_rate = learning_rate).minimize(self.cost)
        
        self.correct_prediction = tf.equal(tf.argmax(self.logits, 1), tf.argmax(self.Y, 1))
        self.accuracy = tf.reduce_mean(tf.cast(self.correct_prediction, "float"))


    

In [4]:

    

learning_rate = 0.01
sess = tf.InteractiveSession()
model = Model(learning_rate, mnist.train.labels.shape[1])
sess.run(tf.global_variables_initializer())


    

    




WARNING:tensorflow:From <ipython-input-3-712396e89d9a>:28: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:

Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.

See tf.nn.softmax_cross_entropy_with_logits_v2.

In [6]:

    

BATCH_SIZE = 128

for i in range(10):
    EPOCH.append(i)
    TOTAL_LOSS, ACCURACY = 0, 0
    for n in range(0, (mnist.train.images.shape[0] // BATCH_SIZE) * BATCH_SIZE, BATCH_SIZE):
        batch_x = mnist.train.images[n: n + BATCH_SIZE, :].reshape((-1, 28, 28, 1))
        cost, _ = sess.run([model.cost, model.optimizer], feed_dict = {model.X : batch_x, model.Y : mnist.train.labels[n: n + BATCH_SIZE, :]})
        ACCURACY += sess.run(model.accuracy, feed_dict = {model.X : batch_x, model.Y : mnist.train.labels[n: n + BATCH_SIZE, :]})
        TOTAL_LOSS += cost
    
    TOTAL_LOSS /= (mnist.train.images.shape[0] // BATCH_SIZE)
    ACCURACY /= (mnist.train.images.shape[0] // BATCH_SIZE)
    LOSS.append(TOTAL_LOSS); ACC.append(ACCURACY)
    print('epoch: %d, avg loss %f, avg acc %f'%(i+1,TOTAL_LOSS,ACCURACY))


    

    




epoch: 1, avg loss 1.035336, avg acc 0.688119
epoch: 2, avg loss 0.227212, avg acc 0.935151
epoch: 3, avg loss 0.159976, avg acc 0.953926
epoch: 4, avg loss 0.130608, avg acc 0.962995
epoch: 5, avg loss 0.111744, avg acc 0.968113
epoch: 6, avg loss 0.098705, avg acc 0.972374
epoch: 7, avg loss 0.089635, avg acc 0.974924
epoch: 8, avg loss 0.082284, avg acc 0.976945
epoch: 9, avg loss 0.076351, avg acc 0.978894
epoch: 10, avg loss 0.071245, avg acc 0.980441

In [7]:

    

model_version = 1
export_model_dir = "./serving/versions"


    

In [8]:

    

import os
export_path_base = export_model_dir
export_path = os.path.join(
            tf.compat.as_bytes(export_path_base),
            tf.compat.as_bytes(str(model_version)))
print('Exporting trained model to', export_path)
builder = tf.saved_model.builder.SavedModelBuilder(export_path)

tensor_input = tf.saved_model.utils.build_tensor_info(model.X)
tensor_output = tf.saved_model.utils.build_tensor_info(model.logits)

prediction_signature = (tf.saved_model.signature_def_utils.build_signature_def(
                inputs={'image': tensor_input},
                outputs={'logits': tensor_output},
                method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME))

builder.add_meta_graph_and_variables(sess, [tf.saved_model.tag_constants.SERVING],
                                     signature_def_map={'predict_classes':prediction_signature,})

builder.save(as_text=True)


    

    




Exporting trained model to b'./serving/versions/1'
INFO:tensorflow:No assets to save.
INFO:tensorflow:No assets to write.
INFO:tensorflow:SavedModel written to: b'./serving/versions/1/saved_model.pbtxt'






    
Out[8]:





b'./serving/versions/1/saved_model.pbtxt'

In [9]:

    

from grpc.beta import implementations
from tensorflow_serving.apis import predict_pb2
from tensorflow_serving.apis import prediction_service_pb2


    

In [13]:

    

server = 'localhost:9000'
host, port = server.split(':')
img=mnist.test.images[0].reshape((28,28,1))
np.expand_dims(img,0).shape


    

    
Out[13]:





(1, 28, 28, 1)

In [12]:

    

channel = implementations.insecure_channel(host, int(port))
stub = prediction_service_pb2.beta_create_PredictionService_stub(channel)


    

In [14]:

    

request = predict_pb2.PredictRequest()
request.model_spec.name = 'mnist'
request.model_spec.signature_name = 'predict_classes'
request.inputs['image'].CopyFrom(
  tf.contrib.util.make_tensor_proto(np.expand_dims(img,0).astype(dtype=np.float32), shape=[1,28,28,1]))


    

In [15]:

    

result_future = stub.Predict(request, 30.)


    

In [17]:

    

np.argmax(result_future.outputs['logits'].float_val)


    

    
Out[17]:





7

In [19]:

    

np.argmax(mnist.test.labels[0])


    

    
Out[19]:





7

In [ ]: