refactor: Skip "best_of_requests" if it is deleted from the installed vLLM

[kthui]

What does the PR do?

Skip "best_of_requests" stats if it is deleted from the installed vLLM package, i.e. >= 0.6.3.

Checklist

• PR title reflects the change and is of format <commit_type>: <Title>
• Changes are described in the pull request.
• Related issues are referenced.
• Populated github labels field
• Added test plan and verified test passes.
• Verified that the PR passes existing CI.
• Verified copyright is correct on all changed files.
• Added succinct git squash message before merging ref.
• All template sections are filled out.
• Optional: Additional screenshots for behavior/output changes with before/after.

Commit Type:

Check the conventional commit type
box here and add the label to the github PR.

• build
• ci
• docs
• feat
• fix
• perf
• refactor
• revert
• style
• test

Related PRs:

#69

Where should the reviewer start?

N/A

Test plan:

This is simply to skip "best_of_requests" stats if it is removed from the installed vLLM package. Existing tests are sufficient at verifying the presence of the stats given the current vLLM version used still contain the stats. A sample output will be provided on the PR verifying the removal of the stats with newer vLLM version.

• CI Pipeline ID:

Caveats:

N/A

Background

best_of_requests stats is removed since vLLM 0.6.3.

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

N/A

[kthui]

On vLLM 0.6.3.post1, where best_of_requests stats is expected to be skipped:

$ curl -s http://localhost:8002/metrics | grep vllm:request_params_
# HELP vllm:request_params_n Histogram of the n request parameter.
# TYPE vllm:request_params_n histogram
vllm:request_params_n_count{model="vllm_model",version="1"} 1
vllm:request_params_n_sum{model="vllm_model",version="1"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="1"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="2"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="5"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="10"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="20"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="+Inf"} 1
$ 

On 0.5.3.post1, where best_of_requests stats is expected to stay:

$ curl -s http://localhost:8002/metrics | grep vllm:request_params_
# HELP vllm:request_params_best_of Histogram of the best_of request parameter.
# TYPE vllm:request_params_best_of histogram
vllm:request_params_best_of_count{model="vllm_model",version="1"} 1
vllm:request_params_best_of_sum{model="vllm_model",version="1"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="1"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="2"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="5"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="10"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="20"} 1
vllm:request_params_best_of_bucket{model="vllm_model",version="1",le="+Inf"} 1
# HELP vllm:request_params_n Histogram of the n request parameter.
# TYPE vllm:request_params_n histogram
vllm:request_params_n_count{model="vllm_model",version="1"} 1
vllm:request_params_n_sum{model="vllm_model",version="1"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="1"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="2"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="5"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="10"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="20"} 1
vllm:request_params_n_bucket{model="vllm_model",version="1",le="+Inf"} 1
$