Skip to content

Commit 55e4ce8

Browse files
authored
fix(agentic): extend warmup drain timeout (#2243)
Increase the AgentX warmup grace period from 10 minutes to 30 minutes so long-context requests can drain before profiling. 中文:将 AgentX 预热排空等待时间从 10 分钟延长至 30 分钟,确保长上下文请求在正式性能采集前完成。
1 parent 6b431f9 commit 55e4ce8

1 file changed

Lines changed: 2 additions & 2 deletions

File tree

benchmarks/benchmark_lib.sh

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1479,9 +1479,9 @@ build_replay_cmd() {
14791479
# with one-token outputs and no idle delays for 10 minutes. Profiling begins
14801480
# only after those requests drain and resumes from the resulting live state.
14811481
REPLAY_CMD+=" --agentic-cache-warmup-duration 600"
1482-
# Give long-context warmup requests up to 10 minutes to drain before
1482+
# Give long-context warmup requests up to 30 minutes to drain before
14831483
# cancelling any remaining requests and starting profiling.
1484-
REPLAY_CMD+=" --warmup-grace-period 600"
1484+
REPLAY_CMD+=" --warmup-grace-period 1800"
14851485
# Use server-reported usage fields (prompt_tokens / completion_tokens) for
14861486
# ISL/OSL instead of client-side tokenizer.encode(). Auto-enables
14871487
# stream_options.include_usage on the OpenAI chat endpoint. Skips the

0 commit comments

Comments
 (0)