[Bugfix] Fix SLA tuner initialization (#27355)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/benchmarks/sweep/serve.py

@@ -66,6 +66,8 @@ def run_benchmark(
 ):
     benchmark_cmd = [
         *bench_overrides.apply_to_cmd(bench_cmd),
+        "--percentile-metrics",
+        "ttft,tpot,itl,e2el",
         "--save-result",
         "--result-dir",
         str(output_path.parent),

vllm/benchmarks/sweep/serve_sla.py

@@ -401,7 +401,8 @@ class SweepServeSLAArgs(SweepServeArgs):
 
     @classmethod
     def from_cli_args(cls, args: argparse.Namespace):
-        base_args = super().from_cli_args(args)
+        # NOTE: Don't use super() as `from_cli_args` calls `cls()`
+        base_args = SweepServeArgs.from_cli_args(args)
 
         if args.sla_params:
             sla_params = SLASweep.read_json(args.sla_params)