topic Re: Attach instance profile to Model serving endpoint in Machine Learning

Attach instance profile to Model serving endpoint

megz — Tue, 29 Aug 2023 18:59:57 GMT

Hi all,

I'm unable to attach an instance profile to a model serving end point. I followed the instructions on this page to update an existing model with an instance profile arn. I have verified the instance profile works by attaching it to a compute cluster. I get the following error in the response:

status code and response.text.

400 {"error_code":"BAD_REQUEST","message":"Instance profile config is not supported!"}

I also tried creating a new end point with instance profile attached. No luck! same error.

Please advise

Thanks!

Re: Attach instance profile to Model serving endpoint

megz — Wed, 30 Aug 2023 16:38:10 GMT

Hi @Retired_mod , thanks for the reply.
I'm not using any cluster for the serving endpoint. I'm using the serverless model deploy option from ML "serving". There's nothing either in the UI or the API that lets me choose the type of cluster.. The only options I have are the model, model version, compute size. I had to attach the instance profile via the API, which gives me the error "Instance profile config is not supported!".

. Thanks.

Re: Attach instance profile to Model serving endpoint

megz — Wed, 30 Aug 2023 16:45:21 GMT

Code snippet in picture attached.
I also tried to update a previously existing serving endpoint (PUT request to /api/2.0/serving-endpoints/{name}/config ) with the new config (which includes instance profile arn) and still get the same error.