gahow/aituner
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Expose nvcc to Q235 FlashInfer profiling

Browse Source
This commit is contained in:
Gahow Wang
2026-07-19 05:42:26 +08:00
parent 39e4719b28
commit 5067bc2cb1
1 changed files with 5 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
runs/frontier-fidelity-envelope-v1/run_q235_after_q30.sh
View File

@@ -10,6 +10,11 @@ RUNNER_DIR="${RUNNER_DIR:-$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)}"
FRONTIER_SOURCE="${FRONTIER_SOURCE:-/home/admin/cpfs/wjh/aituner/frontier-q235-v020-5b953f5}"
VENV_ROOT="${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:${PATH}"
command -v nvcc >/dev/null
mkdir -p "${OUTPUT_ROOT}/supervisor" "${PROFILE_ROOT}"
exec > >(tee -a "${OUTPUT_ROOT}/supervisor/controller.log") 2>&1
echo "Q235_DEFERRED_LAUNCH_ECHO dependency=${Q30_ROOT}:Q30_FIXED_PRESSURE_CAMPAIGN_COMPLETE profile_model=Qwen3-235B-A22B-FP8 profile_backends={TP4/EP1:Triton,TP8/EP8:FlashInfer-CUTLASS} profile_cost=2-7_H20-GPUh experiment_cases={Fixed-PD,Fixed-PO,Trace-PD,Trace-PO} configs={TP4/EP1,TP8/EP8}xMNS{64,128} requests=129 trials=3 expected_campaign_wall=10-30h expected_campaign_cost=90-220_H20-GPUh profile_output=${PROFILE_ROOT} campaign_output=${OUTPUT_ROOT}"