From 909a80f0a66ccf068a947b5649a755b15dfeabaf Mon Sep 17 00:00:00 2001 From: Gahow Wang Date: Sun, 19 Jul 2026 01:13:50 +0800 Subject: [PATCH] Reuse validated FlashInfer cache for pressure probe --- .../fixed-pd-pressure-matching-card.md | 4 +++- .../run_qwen30_fixed_pd_pressure_probe.sh | 6 +++++- 2 files changed, 8 insertions(+), 2 deletions(-) diff --git a/runs/frontier-fidelity-envelope-v1/fixed-pd-pressure-matching-card.md b/runs/frontier-fidelity-envelope-v1/fixed-pd-pressure-matching-card.md index 220cb9a..6b2ec07 100644 --- a/runs/frontier-fidelity-envelope-v1/fixed-pd-pressure-matching-card.md +++ b/runs/frontier-fidelity-envelope-v1/fixed-pd-pressure-matching-card.md @@ -31,7 +31,9 @@ The probe fixes Qwen3-30B-A3B BF16, community vLLM 0.20.0, H20, TP4, MNS64, MBT=8192, chunked prefill, and prefix caching off. It compares per-GPU rates `{1, 2, 3, 4}` (global TP4 rates `{4, 8, 12, 16}` request/s). Every rate has 257 exact `4096 -> 256` requests in each of three fresh-server trials; rate -orders are rotated across trials. +orders are rotated across trials. The launcher reuses the validated vLLM 0.20 +FlashInfer kernel cache; this avoids including one-off custom-kernel JIT in +server startup and does not alter request latency measurement. ## Decision rule diff --git a/runs/frontier-fidelity-envelope-v1/run_qwen30_fixed_pd_pressure_probe.sh b/runs/frontier-fidelity-envelope-v1/run_qwen30_fixed_pd_pressure_probe.sh index 85aed68..c1b4623 100644 --- a/runs/frontier-fidelity-envelope-v1/run_qwen30_fixed_pd_pressure_probe.sh +++ b/runs/frontier-fidelity-envelope-v1/run_qwen30_fixed_pd_pressure_probe.sh @@ -11,7 +11,10 @@ RUNNER_DIR="${RUNNER_DIR:-$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)}" CLIENT="${CLIENT:-${RUNNER_DIR}/../frontier-phase-factorial-v0/qwen30_prefill_client.py}" VENV_ROOT="${VENV_ROOT:-/tmp/wjh/venvs/vllm-0.20.0-cu129-profiler-v1}" MODEL_ROOT="${MODEL_ROOT:-/home/admin/cpfs/wjh/models/Qwen/Qwen3-30B-A3B}" -FLASHINFER_WORKSPACE_BASE="${FLASHINFER_WORKSPACE_BASE:-${OUT}/flashinfer-workspace}" +# Reuse the validated vLLM 0.20 H20 kernel cache. A per-output workspace +# starts costly FlashInfer MoE JIT compilation and changes startup behavior +# without changing the serving configuration being profiled. +FLASHINFER_WORKSPACE_BASE="${FLASHINFER_WORKSPACE_BASE:-/tmp/wjh/flashinfer-workspace-vllm020-profiler-v1}" GPU_IDS="${GPU_IDS:-0,1,2,3}" TP="${TP:-4}" MNS="${MNS:-64}" @@ -211,6 +214,7 @@ PY "${VENV_ROOT}/bin/vllm" --version > "${OUT}/provenance/vllm.version" "${VENV_ROOT}/bin/python" -c 'import torch, transformers, vllm; print(f"torch={torch.__version__}"); print(f"transformers={transformers.__version__}"); print(f"vllm={vllm.__version__}")' > "${OUT}/provenance/runtime.versions" ulimit -n > "${OUT}/provenance/open-file-limit" + readlink -f "${FLASHINFER_WORKSPACE_BASE}" > "${OUT}/provenance/flashinfer-workspace" nvidia-smi --query-gpu=index,name,uuid,driver_version,memory.total --format=csv,noheader > "${OUT}/provenance/gpus.before.csv" declare -a ORDERS=("4 8 12 16" "16 12 8 4" "8 16 4 12")