Build the model
Step 1: Import the necessary libraries
Python3
import tensorflow as tffrom tensorflow import kerasimport numpy as npimport matplotlib.pyplot as pltimport osimport subprocess |
Step 2: Load the dataset
Python3
#Load dataset, dataset is cifar 10. It is open source(data_train, label_train), (data_test, label_test) = keras.datasets.cifar10.load_data()print('Train data :',data_train.shape,label_train.shape)print('Test data :',data_test.shape,label_test.shape) |
Output:
Train data : (50000, 32, 32, 3) (50000, 1)
Test data : (10000, 32, 32, 3) (10000, 1)
Number of classes
Python3
classes = ['aeroplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']num_classes = len(classes)num_classes |
Output:
10
Step 3: Build the model
Python3
model = keras.models.Sequential()model.add(keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))model.add(keras.layers.MaxPooling2D((2, 2)))model.add(keras.layers.Conv2D(64, (3, 3), activation='relu'))model.add(keras.layers.MaxPooling2D((2, 2)))model.add(keras.layers.Conv2D(128, (3, 3), activation='relu'))model.add(keras.layers.Conv2D(128, (3, 3), activation='relu'))model.add(keras.layers.MaxPooling2D((2, 2)))model.add(keras.layers.Flatten())model.add(keras.layers.Dense(256, activation='relu'))model.add(keras.layers.Dense(num_classes, activation='softmax'))# Compile the modelmodel.compile(optimizer='adam', loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=[keras.metrics.SparseCategoricalAccuracy()])# Model summarymodel.summary() |
Output:
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 30, 30, 32) 896
max_pooling2d (MaxPooling2D (None, 15, 15, 32) 0
)
conv2d_1 (Conv2D) (None, 13, 13, 64) 18496
max_pooling2d_1 (MaxPooling (None, 6, 6, 64) 0
2D)
conv2d_2 (Conv2D) (None, 4, 4, 128) 73856
conv2d_3 (Conv2D) (None, 2, 2, 128) 147584
max_pooling2d_2 (MaxPooling (None, 1, 1, 128) 0
2D)
flatten (Flatten) (None, 128) 0
dense (Dense) (None, 256) 33024
dense_1 (Dense) (None, 10) 2570
=================================================================
Total params: 276,426
Trainable params: 276,426
Non-trainable params: 0
_________________________________________________________________
Step 4: Train the model
Python3
epoch = 4model.fit(data_train, label_train, epochs=epoch) |
Output:
Epoch 1/4
1563/1563 [==============================] - 14s 8ms/step - loss: 1.6731 - sparse_categorical_accuracy: 0.4234
Epoch 2/4
1563/1563 [==============================] - 14s 9ms/step - loss: 1.2742 - sparse_categorical_accuracy: 0.5499
Epoch 3/4
1563/1563 [==============================] - 17s 11ms/step - loss: 1.1412 - sparse_categorical_accuracy: 0.5994
Epoch 4/4
1563/1563 [==============================] - 17s 11ms/step - loss: 1.0541 - sparse_categorical_accuracy: 0.6328
<keras.callbacks.History at 0x7f6f5c532920>
Step 5: Evaluate the model
Python3
loss, accuracy_ = model.evaluate(data_test, label_test)print()print(f"Test accuracy: {accuracy_*100}")print(f"Test loss: {loss}") |
Output:
313/313 [==============================] - 1s 4ms/step - loss: 1.0704 - sparse_categorical_accuracy: 0.6358
Test accuracy: 63.58000040054321
Test loss: 1.0704132318496704
Step 6: Save the model
Python3
import tempfileimport os# Specify the current versioncurr_version = 1# Create a temporary directorymy_dir = tempfile.gettempdir()# Create a directory path for the current versionpath = os.path.join(my_dir, str(curr_version))print(f'Export Path = {path}\n')# Saving the current modelmodel.save(path, overwrite=True, include_optimizer=True, save_format=None, signatures=None, options=None)print('\nSaved model:')# Listing the contents of the export pathimport globfile_list = glob.glob(f'{path}/*')for file in file_list: print(file) |
Output:
Export Path = /tmp/1
INFO:tensorflow:Assets written to: /tmp/1/assets
Saved model:
/tmp/1/fingerprint.pb
/tmp/1/saved_model.pb
/tmp/1/assets
/tmp/1/variables
/tmp/1/keras_metadata.pb
Serving a TensorFlow Model
TensorFlow Serving stands as a versatile and high-performance system tailored for serving machine learning models in production settings. Its primary objective is to simplify the deployment of novel algorithms and experiments while maintaining consistent server architecture and APIs. While it seamlessly integrates with TensorFlow models, TensorFlow Serving’s adaptability also enables the service to be expanded for serving diverse model types and data beyond TensorFlow.