Update TF version to 1.11 (aws#442)

* Update TF version to 1.11 * Fixed Keras and Batch Transform notebooks * Fixed Keras Cifar model * Updated TF Pipe Mode Version

Author: mvsusp

hyperparameter_tuning/tensorflow_mnist/hpo_tensorflow_mnist.ipynb

@@ -176,7 +176,7 @@
  "source": [
  "estimator = TensorFlow(entry_point='mnist.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps=1000, \n",
  " evaluation_steps=100,\n",
  " train_instance_count=1,\n",

sagemaker-python-sdk/tensorflow_abalone_age_predictor_using_keras/tensorflow_abalone_age_predictor_using_keras.ipynb

@@ -461,7 +461,7 @@
  "\n",
  "abalone_estimator = TensorFlow(entry_point='abalone.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps= 100, \n",
  " evaluation_steps= 100,\n",
  " hyperparameters={'learning_rate': 0.001},\n",

sagemaker-python-sdk/tensorflow_abalone_age_predictor_using_layers/tensorflow_abalone_age_predictor_using_layers.ipynb

@@ -506,7 +506,7 @@
  "\n",
  "abalone_estimator = TensorFlow(entry_point='abalone.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps= 100, \n",
  " evaluation_steps= 100,\n",
  " hyperparameters={'learning_rate': 0.001},\n",

sagemaker-python-sdk/tensorflow_distributed_mnist/tensorflow_batch_transform_mnist.ipynb

@@ -117,7 +117,7 @@
  "\n",
  "mnist_estimator = TensorFlow(entry_point='mnist.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps=1000, \n",
  " evaluation_steps=100,\n",
  " train_instance_count=2,\n",
@@ -233,11 +233,11 @@
  "outputs": [],
  "source": [
  "import json\n",
- "from urllib.parse import urlparse\n",
+ "from six.moves.urllib import parse\n",
  "\n",
  "import boto3\n",
  "\n",
- "parsed_url = urlparse(transformer.output_path)\n",
+ "parsed_url = parse.urlparse(transformer.output_path)\n",
  "bucket_name = parsed_url.netloc\n",
  "prefix = parsed_url.path[1:]\n",
  "\n",

sagemaker-python-sdk/tensorflow_distributed_mnist/tensorflow_distributed_mnist.ipynb

@@ -149,7 +149,7 @@
  "\n",
  "mnist_estimator = TensorFlow(entry_point='mnist.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps=1000, \n",
  " evaluation_steps=100,\n",
  " train_instance_count=2,\n",

sagemaker-python-sdk/tensorflow_distributed_mnist/tensorflow_local_mode_mnist.ipynb

@@ -165,7 +165,7 @@
  "\n",
  "mnist_estimator = TensorFlow(entry_point='mnist.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " training_steps=10, \n",
  " evaluation_steps=10,\n",
  " train_instance_count=2,\n",

sagemaker-python-sdk/tensorflow_iris_dnn_classifier_using_estimators/tensorflow_iris_dnn_classifier_using_estimators.ipynb

@@ -305,7 +305,7 @@
  "\n",
  "iris_estimator = TensorFlow(entry_point='iris_dnn_classifier.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " output_path=model_artifacts_location,\n",
  " code_location=custom_code_upload_location,\n",
  " train_instance_count=1,\n",

sagemaker-python-sdk/tensorflow_keras_cifar10/cifar10_cnn.py

@@ -18,10 +18,10 @@
 import os
 
 import tensorflow as tf
-from tensorflow.python.keras.layers import InputLayer, Conv2D, Activation, MaxPooling2D, Dropout, Flatten, Dense
+from tensorflow.python.keras.layers import Activation, Conv2D, Dense, Dropout, Flatten, MaxPooling2D
 from tensorflow.python.keras.models import Sequential
-from tensorflow.python.keras.optimizers import RMSprop
 from tensorflow.python.saved_model.signature_constants import PREDICT_INPUTS
+from tensorflow.python.training.rmsprop import RMSPropOptimizer
 
 HEIGHT = 32
 WIDTH = 32
@@ -30,6 +30,7 @@
 NUM_DATA_BATCHES = 5
 NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN = 10000 * NUM_DATA_BATCHES
 BATCH_SIZE = 128
+INPUT_TENSOR_NAME = 'inputs_input' # needs to match the name of the first layer + "_input"
 
 
 def keras_model_fn(hyperparameters):
@@ -44,10 +45,7 @@ def keras_model_fn(hyperparameters):
  """
  model = Sequential()
 
- # TensorFlow Serving default prediction input tensor name is PREDICT_INPUTS. 
- # We must conform to this naming scheme.
- model.add(InputLayer(input_shape=(HEIGHT, WIDTH, DEPTH), name=PREDICT_INPUTS))
- model.add(Conv2D(32, (3, 3), padding='same'))
+ model.add(Conv2D(32, (3, 3), padding='same', name='inputs', input_shape=(HEIGHT, WIDTH, DEPTH)))
  model.add(Activation('relu'))
  model.add(Conv2D(32, (3, 3)))
  model.add(Activation('relu'))
@@ -68,15 +66,13 @@ def keras_model_fn(hyperparameters):
  model.add(Dense(NUM_CLASSES))
  model.add(Activation('softmax'))
 
- _model = tf.keras.Model(inputs=model.input, outputs=model.output)
+ opt = RMSPropOptimizer(learning_rate=hyperparameters['learning_rate'], decay=hyperparameters['decay'])
 
- opt = RMSprop(lr=hyperparameters['learning_rate'], decay=hyperparameters['decay'])
-
- _model.compile(loss='categorical_crossentropy',
+ model.compile(loss='categorical_crossentropy',
  optimizer=opt,
  metrics=['accuracy'])
 
- return _model
+ return model
 
 
 def serving_input_fn(hyperparameters):
@@ -147,7 +143,7 @@ def _input(mode, batch_size, data_dir):
  images, labels = iterator.get_next()
 
  # We must use the default input tensor name PREDICT_INPUTS
- return {PREDICT_INPUTS: images}, labels
+ return {INPUT_TENSOR_NAME: images}, labels
 
 
 def _train_preprocess_fn(image, label):

sagemaker-python-sdk/tensorflow_keras_cifar10/tensorflow_keras_CIFAR10.ipynb

@@ -123,10 +123,18 @@
  "outputs": [],
  "source": [
  "def keras_model_fn(hyperparameters):\n",
+ " \"\"\"keras_model_fn receives hyperparameters from the training job and returns a compiled keras model.\n",
+ " The model will be transformed into a TensorFlow Estimator before training and it will be saved in a \n",
+ " TensorFlow Serving SavedModel at the end of training.\n",
+ "\n",
+ " Args:\n",
+ " hyperparameters: The hyperparameters passed to the SageMaker TrainingJob that runs your TensorFlow \n",
+ " training script.\n",
+ " Returns: A compiled Keras model\n",
+ " \"\"\"\n",
  " model = Sequential()\n",
  "\n",
- " model.add(InputLayer(input_shape=(HEIGHT, WIDTH, DEPTH), name=PREDICT_INPUTS))\n",
- " model.add(Conv2D(32, (3, 3), padding='same'))\n",
+ " model.add(Conv2D(32, (3, 3), padding='same', name='inputs', input_shape=(HEIGHT, WIDTH, DEPTH)))\n",
  " model.add(Activation('relu'))\n",
  " model.add(Conv2D(32, (3, 3)))\n",
  " model.add(Activation('relu'))\n",
@@ -147,15 +155,13 @@
  " model.add(Dense(NUM_CLASSES))\n",
  " model.add(Activation('softmax'))\n",
  " \n",
- " _model = tf.keras.Model(inputs=model.input, outputs=model.output)\n",
- "\n",
- " opt = RMSprop(lr=hyperparameters['learning_rate'], decay=hyperparameters['decay'])\n",
+ " opt = RMSPropOptimizer(learning_rate=hyperparameters['learning_rate'], decay=hyperparameters['decay'])\n",
  "\n",
- " _model.compile(loss='categorical_crossentropy',\n",
+ " model.compile(loss='categorical_crossentropy',\n",
  " optimizer=opt,\n",
  " metrics=['accuracy'])\n",
  "\n",
- " return _model"
+ " return model"
  ]
  },
  {
@@ -216,7 +222,7 @@
  "\n",
  "estimator = TensorFlow(entry_point='cifar10_cnn.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " hyperparameters={'learning_rate': 1e-4, 'decay':1e-6},\n",
  " training_steps=1000, evaluation_steps=100,\n",
  " train_instance_count=1, train_instance_type='ml.c4.xlarge')\n",

sagemaker-python-sdk/tensorflow_pipemode_example/tensorflow_pipemode_example.ipynb

@@ -139,7 +139,7 @@
  "\n",
  "tensorflow = TensorFlow(entry_point='pipemode.py',\n",
  " role=role,\n",
- " framework_version='1.10.0',\n",
+ " framework_version='1.11.0',\n",
  " input_mode='Pipe',\n",
  " output_path=model_artifacts_location,\n",
  " code_location=custom_code_upload_location,\n",