jayinai/serve-ml-with-clipper

Notebook for blog post "Serve Machine Learning Models with Clipper"

The notebook works with Clipper release 0.2

Dependencies

• Install Docker
• Install Python: currently Clipper supports Python 2 only
• Create an Anaconda environment. One straightforward way is to first install conda, and then create the environment by running conda create -n ml-serve python=2.x anaconda, where ml-serve is the name of our Anaconda environment
• Activate the environment: source activate ml-serve
• Install Clipper within the environment: pip install git+https://github.com/ucbrise/clipper.git@develop#subdirectory=clipper_admin

serve-ml-with-clipper's Issues

Exception Handling and Logging

What is the best way to handle errors in a model deployment scenario ? Any pointers would really be helpful !

Notebook loading for long time on deploy the 'predict' function step

Hi @shuaiw

Thanks for sharing this notebook. While testing it, I got stuck at the #Deploy the 'predict' function as a model step where the Jupiter notebook just hanged trying to load. I do understand that under the hood, docker images are being created and indeed I checked it after I ran this step in the notebook.

Is this the reason why the notebook does not stop loading?

Failed to pip sklearn

in this step:
deploy_python_closure(clipper_conn, name="model-1",version=1, input_type="doubles", func=predict,,pkgs_to_install=['numpy==1.14.2','wheel==0.30.0','scripy==1.1.0','sklearn'])

i get this error!
BuildError: The command '/bin/sh -c apt-get -y install build-essential && pip install numpy==1.14.2 wheel==0.30.0 scripy==1.1.0 sklearn' returned a non-zero code: 1

and then i run a simple examples with no pkgs_to_install argument, it successed!
but i run the same simple examples with pkgs_to_install , the same erroer appeared!
can anyone give me a suggestion?

No query frontend

Thanks for the example - and for the explanation accompanying it.

Deploy the 'predict' function as a model

deploy_python_closure(clipper_conn, name="tree-model",
version=1, input_type="doubles", func=predict)

This model is not getting deployed as the error I get is "No Clipper query frontend to attach model container to"
Could you please let me know how to resolve this error ?!