Docs: Inconsistent Model Name in Curl Requests

[danehans]

The quickstart guide currently has separate tabs to test GPU and CPU-based deployment options. The only difference between the two is the model name used in the curl request. Both deployment options should use the same model name.

[danehans]

For those interested in this issue, use the quickstart to create a CPU-based vLLM deployment and use "model": "food-review", instead of "model": "Qwen/Qwen2.5-1.5B-Instruct", to test e2e.

[danehans]

After fixing this issue, the quickstart should not contain any tabs to test e2e since GPU, CPU, and SIM-based deployments use the same model name.

[sathvikpurushotham]

/assign @sathvikpurushotham

[nirrozenbaum]

please note this comment:
#898 (comment)

I'm not 100% sure we want to use the same name as the issue suggests.
we might get additional opinions before pushing a change.

[danehans]

I'm not 100% sure we want to use the same name as the issue suggests.

@nirrozenbaum I updated the issue title to reduce confusion. This issue is about using the same model name, food-review, in the curl request body for all quickstart vLLM deployment options. The food-review InferenceModel will route model: food-review requests to model servers regardless of the base model differences between CPU and non-CPU vLLM deployments.

[linsun]

+1, also run into this myself, and started a conversation on slack if anyone is interested: https://kubernetes.slack.com/archives/C08E3RZMT2P/p1753119528560539