From b80fab815dc802365d4642bcc001b451f19afee9 Mon Sep 17 00:00:00 2001 From: Sugam Devare Date: Fri, 4 Sep 2026 21:45:13 +0000 Subject: [PATCH] fix(nemotron_3.5_super): repin vllm-router to the current head of #238 The pin added in #3080 pointed at a commit that also carried per-role balance threshold flags (--prefill-balance-abs-threshold and friends). Those were an escape hatch we never needed: they default to the existing global values, so they are inert unless passed, and nothing here passes them. They have been dropped from vllm-project/router#238 to keep the PR to the bug fix, so this moves the pin to the new head. No behaviour change. The remaining commits are unchanged: #216 (health checker no longer resets in-flight worker load) plus the fix for the prefill-side regression #216 exposes, where cache_aware decided prefix affinity from the fleet-wide load spread instead of per request. Repinning also removes an incidental dependency: the previously pinned commit is only reachable through a contributor branch, whereas the new one is the current PR head. Signed-off-by: Sugam Devare --- benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh b/benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh index 7067be2f55..e2cb344e54 100755 --- a/benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh +++ b/benchmarks/nemotron_3.5_super/build_vllm_router_wheel.sh @@ -34,7 +34,7 @@ VLLM_ROUTER_GIT_URL=${VLLM_ROUTER_GIT_URL:-https://github.com/vllm-project/route # to preserve worker load during checks") unchanged plus a fix for the prefill-side # regression #216 exposes; see the README. Fetching a bare SHA works because GitHub # serves any commit reachable from a ref, and PR heads are refs. -VLLM_ROUTER_COMMIT=${VLLM_ROUTER_COMMIT:-5b7ce08ea8cd27123cf6c23313cd92d3ba0da70e} +VLLM_ROUTER_COMMIT=${VLLM_ROUTER_COMMIT:-84b4425d7550b993d17010db0408ab170dc6eecb} RUST_TOOLCHAIN=${RUST_TOOLCHAIN:-1.95.0} # Load-accounting unit tests from the fix. Compiling the test binary roughly # doubles the job, so it is opt-out.