Error adding endpoint to TabPy Server on Google Cloud VM Instance

Hi,

We've been trying to get a TabPy server running on a Google Cloud VM so that we can deploy models from Datalab. I've installed TabPy and the server is running fine - navigating to the ip address in a browser brings up the Tableau logo, and I'm able to connect Tableau to the predictive service and execute python code using the SCRIPT_ functions - however I'm getting an error when using the client to deploy from Datalab.

Using the code:

import tabpy_client
client = tabpy_client.Client('http://{0}:9004/').format(server_ip)

def add(x,y):
    import numpy as np
    return np.add(x, y).tolist()

client.deploy('add', add, 'Adds two numbers x and y')

Brings up the error:

I may well have overlooked something in getting it set up as it is working fine locally, but can't work out what's causing it.

Thanks in advance,

Adam

[BBeran]

Currently TabPy only supports deploying to local file paths. So this is not a bug, it is a feature request :)

[Seiter02]

This makes deploying to TabPy server running in docker containers impossible as well. Feature request +1

[afreeland]

Believe I am running into this with a docker image as well. Does this mean that if we
pip install tabpy-client future from an integrated terminal inside the tabpy server container that it should work? Not quite sure in what context local needs to be (server or client)...

I tried, with the example from above directly in a file in the container

import tabpy_client
client = tabpy_client.Client('http://localhost:9004/')

def add(x,y):
    import numpy as np
    return np.add(x, y).tolist()

client.deploy('add', add, 'Adds two numbers x and y')

However, I get an error

RuntimeError: LoadFailed: u"Load failed: 'int' object is not iterable"

[Seiter02]

Yeah, it will work inside the container if you mount the file from the host.

The error you are seeing with the number not being iterable is something else unrelated to docker or localhost clients, but more likely related to the versions of python/numpy you are using in your image.

[afreeland]

@Seiter02 I am new to TabPy, can you provide a sample of what file needs to be mounted in the host? When I was looking the staging folder and such never existed.

Using this version of python
Python 2.7.12 :: Continuum Analytics, Inc

Looks like numpy wasnt installed...but now it is

pip install numpy
python -c "import numpy; print(numpy.version.version)"
# 1.14.1

Still same error, will continue looking through it

Using this Docker Image and this Dockerfile

[nmannheimer]

Hi @afreeland,

Are you still experiencing this issue?

[afreeland]

@nmannheimer I ended up not continuing with TabPy because we had a large datasets and Tableau executes the script calculated field requests synchronously which didnt allow us to scale the way we needed...sorry can't help much

[nmannheimer]

@afreeland thanks for the information, that's useful for us to know.