[Bug]: ovms-gpu-ootb is missing recommended accelerator annotation

[Gkrumbach07]

Is there an existing issue for this?

• I have searched the existing issues

Deploy type

OpenDataHub core version (eg. v1.6.0)

Version

v2.16.0

Current Behavior

the OOTB gpu model serving template is missing the recommended accelerators annotation for nvidia.com/gpu so the label does not show up when selecting a NVIDIA GPU.

Notes: The OOTB GPU enabled model serving template is missing the annotation 'opendatahub.io/recommended-accelerators': '["nvidia.com/gpu"]' which makes the "Recommended serving runtime" label not show up in the "Add model server" modal's Serving Runtime dropdown options.

Expected Behavior

for the label to show up when selecting the NVIDIA GPU

Steps To Reproduce

1. open model server create modal
2. select the gpu model server
3. accelerators with identifier = nvidia.com/gpu wont have a recommended label next to it

Workaround (if any)

not a breaking bug, but you can manually add the label when making a copy of the template model server

What browsers are you seeing the problem on?

No response

Anything else

No response

[bredamc]

@Gkrumbach07 If you would like to include this issue in the 1.34 Release Notes, please provide the text for the "Known issues" section.

[Gkrumbach07]

@bredamc here is something

The OOTB GPU enabled model serving template is missing the annotation opendatahub.io/recommended-accelerators: '["nvidia.com/gpu"]' which makes the "Recommended serving runtime" label not show up in the "Add model server" modal's Serving Runtime dropdown options, when it should be there.

[lucferbux]

Closing this cause we are going to fix it in #1989

[bredamc]

@andrewballantyne Do you still want to include this bug in the Known issues section of the 1.34 Release Notes?

If so, is the following text ok?

"
ODH-DASHBOARD-1991 - ovms-gpu-ootb is missing recommended accelerator annotation

When you add a model server to your project, the Serving runtime list does not show the "Recommended serving runtime" label for the NVIDIA GPU.

Workaround
Make a copy of the model-server template and manually add the label.
"

[andrewballantyne]

@bredamc Yup that works... it's unfortunate but should be fixed in 1.36 when we release KServe