qqtang/sskj
1
3
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[Fix] Persist FlashInfer JIT cache across services

Browse Source
This commit is contained in:
Zhiyi Hong 2026-08-19 11:16:07 +08:00
parent 7f67dfe6b3
commit ab9a5422f6
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
experiments/pro6000/kimi3_pro6000_sglang_sm120_flashinfer_mxfp4/run_pr_representative_point.sh
View File

@ -160,6 +160,7 @@ prewarm_one_node() {
sudo_host "$host" mkdir -p "${FLASHINFER_CACHE}" sudo_host "$host" mkdir -p "${FLASHINFER_CACHE}"
sudo_host "$host" docker run --rm --gpus device=0 \ sudo_host "$host" docker run --rm --gpus device=0 \
-v "${FLASHINFER_CACHE}:/root/.cache/flashinfer" \ -v "${FLASHINFER_CACHE}:/root/.cache/flashinfer" \
-e FLASHINFER_WORKSPACE_BASE=/root \
-e FLASHINFER_DISABLE_VERSION_CHECK=1 \ -e FLASHINFER_DISABLE_VERSION_CHECK=1 \
--entrypoint python3 "${PR_IMAGE}" -m pytest -q -s \ --entrypoint python3 "${PR_IMAGE}" -m pytest -q -s \
/opt/pr-tests/test_mxfp4_sm120_cutlass.py \ /opt/pr-tests/test_mxfp4_sm120_cutlass.py \
@ -238,6 +239,7 @@ start_node() {
-e NCCL_IB_GID_INDEX=3 -e NCCL_IB_TIMEOUT=22 -e NCCL_IB_RETRY_CNT=7 -e NCCL_IB_GID_INDEX=3 -e NCCL_IB_TIMEOUT=22 -e NCCL_IB_RETRY_CNT=7
-e NCCL_CUMEM_ENABLE=1 -e PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True -e NCCL_CUMEM_ENABLE=1 -e PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
-e SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0 -e SGLANG_MOE_FUSED_GATE_RADIX=1 -e SGLANG_ENABLE_TP_MEMORY_INBALANCE_CHECK=0 -e SGLANG_MOE_FUSED_GATE_RADIX=1
-e FLASHINFER_WORKSPACE_BASE=/root
-e FLASHINFER_DISABLE_VERSION_CHECK=1 -e FLASHINFER_DISABLE_VERSION_CHECK=1
"${PR_IMAGE}" -lc "$bootstrap" "${PR_IMAGE}" -lc "$bootstrap"
) )