Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@validators/performance/inference_perf_constraint.go`:
- Around line 1234-1253: The buildAIPerfJob doc comment is now separated from
its declaration by the newly inserted intFromEnv function; move the existing doc
comment block that describes buildAIPerfJob so it immediately precedes the func
buildAIPerfJob declaration (or alternatively move intFromEnv below
buildAIPerfJob) so the Go doc comment is attached to buildAIPerfJob; ensure the
comment stays unchanged and sits directly above func buildAIPerfJob.

In `@validators/performance/inference_perf_test.go`:
- Around line 597-609: The test TestBuildAIPerfJob_Warmup should be refactored
into a table-driven style: create a slice of test cases (each with a name and
concurrency value), iterate over them and run each as t.Run(tc.name, func(t
*testing.T) { ... }), calling buildAIPerfJob and asserting the script contains
the expected "--warmup-request-count" needle; keep using the existing
helpers/variables like buildAIPerfJob and aiperfWarmupPerConcurrency and
preserve the current assertion message but move it inside each subtest.

---

Outside diff comments:
In `@validators/performance/inference_perf_constraint.go`:
- Around line 1270-1291: The shell script built in the fmt.Sprintf assigned to
variable "script" does not quote the URL flag (--url %s); update the format
string to pass the URL quoted (e.g., --url "%s") so the generated command uses
--url "<endpoint>" just like the model name is quoted in aiperf profile "%s";
ensure the corresponding argument ordering (endpoint) remains unchanged in the
call to fmt.Sprintf in validators/performance/inference_perf_constraint.go.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@validators/performance/inference_perf_constraint.go`:
- Around line 1241-1253: The intFromEnv helper currently swallows malformed
AICR_INFERENCE_PERF_* env values and returns a default; change it to surface
parse errors by returning an error so the caller can abort with
ErrCodeInvalidRequest. Update intFromEnv(name string, def int) to return (int,
error), parse the env and on empty return (def, nil), but on strconv.Atoi error
or v <= 0 return (0, fmt.Errorf(...)) (or a typed sentinel error); then update
the callers (e.g., where intFromEnv is used) to check the error and return
ErrCodeInvalidRequest so the process aborts rather than silently using defaults.
Ensure references to ErrCodeInvalidRequest are used for the abort path and
propagate the new error signature through the call chain.

---

Outside diff comments:
In `@validators/performance/inference_perf_constraint.go`:
- Around line 1256-1287: The code builds the aiperf command from derived knobs
(requestCount, warmupCount, input/output token means, concurrency, aiperfImage)
but only logs the constant aiperfMinRequests; update runAIPerfJob to emit a
debug/info log right after resolving aiperfImage (or before creating script)
that prints the computed values requestCount, warmupCount, concurrency,
intFromEnv(envInputTokensMean,...), intFromEnv(envOutputTokensMean,...), and the
resolved aiperfImage so the run artifacts reflect the actual knobs used; place
the log near the existing script construction (referencing variables
requestCount, warmupCount, concurrency, aiperfImage) and use the project's
standard logger (e.g., processLogger or the logger used elsewhere in
runAIPerfJob) to ensure it appears in job outputs.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@docs/contributor/validator.md`:
- Line 152: Update the subsection text to expand the acronym TTFT on its first
mention by replacing the first occurrence of "TTFT" with "time-to-first-token
(TTFT)"; ensure the rest of the paragraph continues to use "TTFT" thereafter and
do not change identifiers like "inference-perf", the "env" knobs, or the error
symbol "ErrCodeInvalidRequest".

In `@validators/performance/inference_perf_test.go`:
- Around line 651-653: Update the top-of-test comment for TestIntFromEnv to
accurately describe the behavior: state that a valid positive integer is
accepted, an unset value falls back to the default, and non-integer, zero, or
negative values are rejected (fail-closed) with an error; mention the
TestIntFromEnv test and the env-reader function it exercises so readers
understand it's enforcing error-on-invalid rather than silently defaulting.