[Benchmark] Record model_id in bench latency/throughput --output-json (#58112)
Signed-off-by: yashasvi <yashasvi@ibm.com>
Yashasvi Chaurasia committed
Sep 25, 2026 at 22:13 UTC
f83ea61cfe77f0b200f2ded1fda9be5ff591af2c
2 files changed
+2
vllm/benchmarks/latency.py
+1
@@ -169,6 +169,7 @@ def main(args: argparse.Namespace):
169
# Output JSON results if specified
170
if args.output_json:
171
results = {
172
+ "model_id": args.model,
173
"avg_latency": np.mean(latencies),
174
"latencies": latencies.tolist(),
175
"percentiles": dict(zip(percentages, percentiles.tolist())),
vllm/benchmarks/throughput.py
+1
@@ -1141,6 +1141,7 @@ def main(args: argparse.Namespace):
1141
# Output JSON results if specified
1142
if args.output_json:
1143
results = {
1144
+ "model_id": args.model,
1145
"elapsed_time": elapsed_time,
1146
"num_requests": len(requests),
1147
"total_num_tokens": total_num_tokens,