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 f96b5b8..944813b 100644 --- a/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh +++ b/runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh @@ -14,6 +14,7 @@ VENV_ROOT="${VENV_ROOT:-/tmp/wjh/venvs/vllm-0.20.0-cu129-profiler-v1}" # autotunes a previously unseen FP8 grouped-GEMM shape. export PATH="/usr/local/cuda/bin:${PATH}" export FRONTIER_VLLM_MODEL_TYPE=qwen3_moe +export VLLM_KV_CACHE_LAYOUT=NHD command -v nvcc >/dev/null mkdir -p "${OUTPUT_ROOT}/supervisor" "${PROFILE_ROOT}"