Python: Advanced Guide to Artificial Intelligence

Serving the model in the Docker container

To serve the model in the container, the instructions are as follows:

1. Start the MNIST container built in the previous section:

$ docker run --name=mnist_container -it $USER/mnist_serving

2. Give the cd command to go to the home folder.
3. Run the ModelServer with the following command:

$ tensorflow_model_server  --model_name=mnist --model_base_path=/tmp/mnist_model/ &> mnist_log &

4. Check the prediction from the model with the sample client:

$ python serving/tensorflow_serving/example/mnist_client.py --num_tests=100 --server=localhost:8500

5. We see the error rate to be 7%, just like our previous notebook example execution:

Extracting /tmp/train-images-idx3-ubyte.gzExtracting /tmp/train-labels-idx1-ubyte.gz ...