From d563c30b422b6f71dcc75bcbfe99ffd538e205e8 Mon Sep 17 00:00:00 2001
From: Gahow Wang <gahow.wang@gmail.com>
Date: Sun, 19 Jul 2026 06:12:11 +0800
Subject: [PATCH] Set Qwen3 model type for vLLM profile dispatch

---
 runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh | 1 +
 1 file changed, 1 insertion(+)

diff --git a/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh b/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh
index 90fe528..f96b5b8 100644
--- a/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh
+++ b/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh
@@ -13,6 +13,7 @@ VENV_ROOT="${VENV_ROOT:-/tmp/wjh/venvs/vllm-0.20.0-cu129-profiler-v1}"
 # FlashInfer's TensorRT-LLM MoE runtime invokes `nvcc` by name when it
 # autotunes a previously unseen FP8 grouped-GEMM shape.
 export PATH="/usr/local/cuda/bin:${PATH}"
+export FRONTIER_VLLM_MODEL_TYPE=qwen3_moe
 command -v nvcc >/dev/null
 
 mkdir -p "${OUTPUT_ROOT}/supervisor" "${PROFILE_ROOT}"