ML Foundation Bring your own Model

[Former Member]

Hi,

I have a Tensorflow saved Model in .pb Format which is packed in a .zip and want to deploy it in my SCP Cloud with Cloud Foundry Trial.

There is a Blog Article and a Documentation which are a good help.

But when I am deploying my model and check the status of the Model after a long Pending status I get:

"deploymentStatus": {
                "state": "FAILED",
                "description": "Deployment could not be completed due to a timeout. Please try again after some time or contact administrator"
            },

Well that's really hard to debug, without any other Information, so maybe someone also got experience with it?

[Former Member · Accepted solution]

I solved the Problem,

the deployed Model has to be in the savedModel Format, and needs to be zipped. I just tried it with a .pb file which is not enough.

If you're using Keras to create the model, use this code here https://gist.github.com/asimshankar/000b8d276f211f972168afa138eb3cc7 to transform the model in the right format and the deployment should work fine.