A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 11, 17:30 UTC
[Misc] Consistent case for `vllm bench serve` results (#30403)
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Matthew Bonanni committed
Dec 10, 2025 at 12:44 UTC
794a7875ee0df7d2c12ff0ba83b76438ca68bf26
3 files changed
+3
-3
benchmarks/benchmark_serving_structured_output.py
+1
-1
@@ -574,7 +574,7 @@ async def benchmark(
574
)
575
print(
576
"{:<40} {:<10.2f}".format(
577
- "Total Token throughput (tok/s):", metrics.total_token_throughput
577
+ "Total token throughput (tok/s):", metrics.total_token_throughput
578
)
579
)
580
docs/benchmarking/cli.md
+1
-1
@@ -84,7 +84,7 @@ Total input tokens: 1369
84
Total generated tokens: 2212
85
Request throughput (req/s): 1.73
86
Output token throughput (tok/s): 382.89
87
-Total Token throughput (tok/s): 619.85
87
+Total token throughput (tok/s): 619.85
88
---------------Time to First Token----------------
89
Mean TTFT (ms): 71.54
90
Median TTFT (ms): 73.88
vllm/benchmarks/serve.py
+1
-1
@@ -788,7 +788,7 @@ async def benchmark(
788
)
789
print(
790
"{:<40} {:<10.2f}".format(
791
- "Total Token throughput (tok/s):", metrics.total_token_throughput
791
+ "Total token throughput (tok/s):", metrics.total_token_throughput
792
)
793
)
794