TensorFlow Examples

Setup TensorFlow

cd $MYAPPS_HOME/tensorflow
./setup-tensorflow.sh
...
No GPU support will be enabled for TensorFlow
Configuration finished

Setup TensorFlow Serving (Requires 24 GB RAM within Docker Container)

cd $MYAPPS_HOME/serving/tensorflow
./setup-tensorflow-serving.sh
...
No GPU support will be enabled for TensorFlow
Configuration finished

Train and Deploy v1 MNIST Model to TensorFlow Serving

• This will train and deploy the v1 model with 1,000 training iterations
• TensorFlow Serving will automatically pick up the v1 MNIST Model
• This step will take a while (3-5 mins), so please be patient

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/mnist_export --training_iteration=1000 --export_version=1 $DATASETS_HOME/tensorflow/serving/mnist_model

Start TensorFlow Serving with MNIST Model (Port 9090)

• This will pick up the v1 MNIST Model that has been training using 1,000 training iterations

cd $MYAPPS_HOME/tensorflow
./start-tensorflow-mnist-serving-service.sh

• Verify that TensorFlow Serving has found the v1 MNIST Model mnist_model/00000001

tail -f $LOGS_HOME/tensorflow/serving/nohup-mnist.out 
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:85] Found a servable {name: default version: 1} at path /root/pipeline/datasets/tensorflow/serving/mnist_model/00000001
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:149] Aspiring 1 versions for servable default

• ctrl-c to exit the tail log

Perform MNIST Classification using v1 MNIST Model (Port 9090)

• This will run 1000 tests against the v1 MNIST Model and output an Inference error rate metric that we'll try to minimize in the next step
• (This is kind of boring, but this will get more interesting down below... stick with us.)

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/mnist_client --num_tests=1000 --server=localhost:9090
...
Inference error rate: 13.1%

Train and Deploy v2 MNIST Model to TensorFlow Serving

• This will train and deploy the v2 model with 10,000 training iterations - an order of magnitude more than v1
• TensorFlow Serving will automatically pick up the new v2 MNIST Model
• This step will take a while (3-5 mins), so please be patient

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/mnist_export --training_iteration=10000 --export_version=2 $DATASETS_HOME/tensorflow/serving/mnist_model

• Verify that TensorFlow Serving has found the new v2 MNIST Model mnist_model/00000002

tail -f $LOGS_HOME/tensorflow/serving/nohup-mnist.out 
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:149] Aspiring 1 versions for servable default
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:45] Polling the file system to look for a servable path under: /root/pipeline/datasets/tensorflow/serving/mnist_model
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:85] Found a servable {name: default version: 2} at path /root/pipeline/datasets/tensorflow/serving/mnist_model/00000002

• ctrl-c to exit the tail log

Perform MNIST Classification using v2 MNIST Model (Port 9090)

• This will run the same number of tests (1000) as the prior run, but against the new v2 MNIST Model
• Inference error rate should have decreased from the earlier run

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/mnist_client --num_tests=1000 --server=localhost:9090
...
Inference error rate: 9.5%

Start TensorFlow Serving with Inception Model (Port 9091)

• We've already deployed the model to $DATASETS_HOME/tensorflow/serving/inception_model to be picked up by TensorFlow Serving
• We just need to start the TensorFlow Serving Classification Service as follows:

$MYAPPS_HOME/serving/tensorflow/start-tensorflow-inception-serving-service.sh

• Verify that TensorFlow Serving found the v00157585 Inception Model at /root/pipeline/datasets/tensorflow/serving/inception_model/00157585

tail -f $LOGS_HOME/tensorflow/serving/nohup-inception.out
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:149] Aspiring 1 versions for servable default
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:45] Polling the file system to look for a servable path under: /root/pipeline/datasets/tensorflow/serving/inception_model
tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:85] Found a servable {name: default version: 157585} at path /root/pipeline/datasets/tensorflow/serving/inception_model/00157585

Perform Image Classification using Inception Model (Port 9091)

• This is an example client that uses the gRPC-based TensorFlow Serving Image Classification Service to classify an image

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/inception_client --server=localhost:9091 --image=$DATASETS_HOME/inception/cropped_panda.jpg

Classify Your Own Image using Inception Model (Port 9091)

wget <my-remote-image-url>

cd $TENSORFLOW_SERVING_HOME

$TENSORFLOW_SERVING_HOME/bazel-bin/tensorflow_serving/example/inception_client --server=localhost:9091 --image=<my-local-image-filename>

Image Classification Web Service

• Classify an image passed in as image_url

http://<ip>:5070/classify?image_url=https://bdproffer.files.wordpress.com/2014/04/cropped-panda.jpg

• TODO: Accept uploaded image

Classify an Image at Full Precision

bazel-bin/tensorflow/examples/label_image/label_image \
  --graph=$DATASETS_HOME/inception/classify_image_graph_def.pb \
  --input_width=299 \
  --input_height=299 \  
  --input_mean=128 \  
  --input_std=128 \  
  --input_layer="Mul:0" \
  --output_layer="softmax:0" \
  --image=$DATASETS_HOME/inception/cropped_panda.jpg

...

W tensorflow/core/framework/op_def_util.cc:332] Op BatchNormWithGlobalNormalization is deprecated. It will cease to work in GraphDef version 9. Use tf.nn.batch_normalization().
I tensorflow/examples/label_image/main.cc:207] giant panda (169): 0.89233
I tensorflow/examples/label_image/main.cc:207] indri (75): 0.00858706
I tensorflow/examples/label_image/main.cc:207] lesser panda (7): 0.00264235
I tensorflow/examples/label_image/main.cc:207] custard apple (325): 0.00140671
I tensorflow/examples/label_image/main.cc:207] earthstar (878): 0.00107063

Classify an Image after 8-bit Quantization

• http://www.kdnuggets.com/2016/05/how-quantize-neural-networks-tensorflow.html
• This will produce a new model that runs the same operations as the original, but with eight bit calculations internally, and all weights quantized as well.
• If you look at the file size, you’ll see it’s about a quarter of the original (23MB versus 91MB).
• You can still run this model using exactly the same inputs and outputs though, and you should get equivalent results.
• Here's an example:

# As of May 2016, you must replace the existing `label_image/BUILD` file
#   with the one from our repo in order to pick up the quantization libs
mv $TENSORFLOW_HOME/tensorflow/examples/label_image/BUILD $TENSORFLOW_HOME/tensorflow/examples/label_image/BUILD.orig

cp $MYAPPS_HOME/tensorflow/hack-add-quantization-dependencies-BUILD $TENSORFLOW_HOME/tensorflow/examples/label_image/BUILD

# Build the label_image example
bazel build tensorflow/examples/label_image/...

# Classify the image using the smaller (1/3rd) quantization model 
#   and compare results to full-precision model above (equivalent)
bazel-bin/tensorflow/examples/label_image/label_image \
 --graph=$DATASETS_HOME/inception/quantized_classify_image_graph_def.pb \
 --input_width=299 \
 --input_height=299 \ 
 --input_mean=128 \
 --input_std=128 \
 --input_layer="Mul:0" \
 --output_layer="softmax:0" \
 --image=$DATASETS_HOME/inception/cropped_panda.jpg
...

tensorflow/examples/label_image/main.cc:207] giant panda (169): 0.849015
tensorflow/examples/label_image/main.cc:207] indri (75): 0.00533261
tensorflow/examples/label_image/main.cc:207] lesser panda (7): 0.00349496
tensorflow/examples/label_image/main.cc:207] custard apple (325): 0.00193439
tensorflow/examples/label_image/main.cc:207] earthstar (878): 0.0013225

8-bit Quantization of Inception Model

• This has already been done for you, but if you would like to do it manually, here's how:

curl http://download.tensorflow.org/models/image/imagenet/inception-2015-12-05.tgz -o /tmp/inceptionv3.tgz

bazel build tensorflow/contrib/quantization/tools:quantize_graph
bazel build //tensorflow/contrib/quantization/tools:quantize_graph

bazel-bin/tensorflow/contrib/quantization/tools/quantize_graph \
  --input=$DATASETS_HOME/inception/classify_image_graph_def.pb \
  --mode=eightbit \
  --output_node_names=softmax2 \
  --output=$DATASETS_HOME/inception/quantized_classify_image_graph_def.pb

TODO: SyntaxNet

• https://github.com/tensorflow/models/tree/master/syntaxnet

TODO: NLP Sentence Prediction with PTB

• Build Code

cd $TENSORFLOW_HOME
bazel build //tensorflow/models/rnn/ptb:ptb_word_lm

• Train Model

python ptb_word_lm.py --data_path=$DATASETS_HOME/ptb/simple-examples/data --model small

Predict Sentence

TODO

Troubleshooting

src/main/tools/namespace-sandbox.c:774: mount("none", "/", NULL, MS_REC | MS_PRIVATE, NULL): Permission denied

• You're likely running this inside a Docker Container that has not been run with --privileged