minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Revert "fix: use consistent time denominator for throughput metrics in bench_one_batch_server" (#21276)

Browse Source
This commit is contained in:
Cheng Wan
2026-03-23 22:14:54 -07:00
committed by GitHub
parent 3837588a4c
commit c01ee848b0
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/test/bench_one_batch_server_internal.py
+2 -2
View File
@@ -600,8 +600,8 @@ def run_one_case(
# Compute metrics
latency = time.perf_counter() - tic
input_throughput = batch_size * input_len / latency
output_throughput = batch_size * output_len / latency
input_throughput = batch_size * input_len / last_ttft
output_throughput = batch_size * output_len / (latency - last_ttft)
overall_throughput = batch_size * (input_len + output_len) / latency
if backend == "vllm":