Support config GPU vendor

[typhoonzero]

This is a following up work of #2973.

Removed "vgpu memory" boxes.

What changes were proposed in this pull request?

The K8s cluster used for GPU training may have other GPU resource types other than nvidia.com/gpu, there are other vendors like AMD or opensource vGPU implementations, like https://github.com/4paradigm/k8s-vgpu-scheduler, https://github.com/tkestack/gpu-manager etc. I propose to add support for setup user inputed resource types.

How was this pull request tested?

Tests added at test_pipeline_constructor.py

[elyra-bot]

Thanks for making a pull request to Elyra!

To try out this branch on binder, follow this link:

[ptitzler]

If the user enters an invalid gpu_vendor value (e.g. only spaces ' ' or a string that violates Kubernetes naming constraints) pod execution will fail:

We therefore also need to add a few lines to the existing validation code to prevent these failures.

I checked our utility library and it appears that function is_valid_annotation_key performs the kind of checking we need. We could therefore create a wrapper function like this one and invoke it after this code.

[typhoonzero]

@ptitzler Thanks for the detailed comments, the PR is updated. By the way, I'm looking for a way to put some UI components under an "advanced options" section, which can be toggled by the user. I've noticed that there are such components used at the left bar for metadata UI, is there such component in the right bar and how to apply it?

[ptitzler]

@ptitzler Thanks for the detailed comments, the PR is updated. By the way, I'm looking for a way to put some UI components under an "advanced options" section, which can be toggled by the user. I've noticed that there are such components used at the left bar for metadata UI, is there such component in the right bar and how to apply it?

Unfortunately not yet. We've discussed this in our dev meeting a while ago and have identified a potential solution that would enable the Pipeline Editor to distinguish between core properties (that are always visible, e.g. because they are required) and properties that are only of relevance to advanced users. This solution should be flexible enough to enable users to switch between basic display mode and advanced mode. This should help improve the up-and-running experience, as there is currently a fairly large number of [Kubernetes] properties that can be intimidating at first.

[ptitzler]

The changes are looking good!! Before we merge, can you please update the documentation topic at https://github.com/elyra-ai/elyra/blob/main/docs/source/user_guide/pipelines.md#resources-cpu-gpu-and-ram to mention the new property and perhaps link to https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/ (or another URL that might be useful to the user.) Thank you!

because

[ptitzler]

LGTM! Thank you for your contribution! The functionality will be included in Elyra 3.14.

(I tweaked the documentation a tiny bit. For example, I reversed the proposed header change because it broke existing references to the topic.)