[RFC] unified benchmark script for all examples under GenAIExamples

[ftian1]

This RFC is used to demonstrate the enhanced design of GenAIExamples benchmark script.

[AlexHe99]

Is that make sense to set specific GPU by [instance_num] if using k8s? It is usual to set the number of GPU to k8s and k8s will auto manager the GPU resource.

[ftian1]

Is that make sense to set specific GPU by [instance_num] if using k8s? It is usual to set the number of GPU to k8s and k8s will auto manager the GPU resource.

it's used to provide a manual control way for user to specify the concrete config. of course we should also support auto-scale way in k8s by ignoring this field.

[poussa]

Why the ´ deploy_and_benchmark` file is in the root?

[mkbhanda]

Why the ´ deploy_and_benchmark` file is in the root?

Because it works across all the GenAIExamples. It takes as argument the chatname.
As name suggestes, it deploys the workload specified by chartname and runs the benchmark (performance and/or accuracy) specified in the default or provided yaml file.