fix(build): 所有子脚本去掉set -euo pipefail + 全面容错
- install_prebuilt_corex.sh: set -euo pipefail → set +e, exit 2 → 非致命warning - build_moe_topk.sh: set -euo pipefail → set +e - patch_ops.sh: install_prebuilt_corex.sh 调用加 || echo non-fatal 26e6cb40没有这些子脚本。新增的子脚本用了set -euo pipefail会在 竞赛平台环境差异下(无GPU/权限不同/路径不同)触发exit非零, 虽然patch_ops.sh没set -e不会退出,但子进程的strict模式 可能导致意外的级联失败。
This commit is contained in:
@@ -1,6 +1,6 @@
|
|||||||
#!/usr/bin/env bash
|
#!/usr/bin/env bash
|
||||||
# build_moe_topk.sh — Compile moe_topk_softmax_v3.cu into importable .so
|
# build_moe_topk.sh — Compile moe_topk_softmax_v3.cu into importable .so
|
||||||
set -euo pipefail
|
set +e
|
||||||
cd "$(dirname "$0")"
|
cd "$(dirname "$0")"
|
||||||
|
|
||||||
PYTHON=${PYTHON:-python3}
|
PYTHON=${PYTHON:-python3}
|
||||||
|
|||||||
@@ -1,5 +1,6 @@
|
|||||||
#!/usr/bin/env bash
|
#!/usr/bin/env bash
|
||||||
set -euo pipefail
|
# Tolerant: any failure is non-fatal
|
||||||
|
set +e
|
||||||
|
|
||||||
VLLM_ROOT=${1:?usage: install_prebuilt_corex.sh VLLM_ROOT}
|
VLLM_ROOT=${1:?usage: install_prebuilt_corex.sh VLLM_ROOT}
|
||||||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||||||
@@ -8,24 +9,24 @@ MANIFEST=${BUNDLE_DIR}/SHA256SUMS
|
|||||||
|
|
||||||
[[ -d "$VLLM_ROOT" ]] || {
|
[[ -d "$VLLM_ROOT" ]] || {
|
||||||
printf 'vLLM root does not exist: %s\n' "$VLLM_ROOT" >&2
|
printf 'vLLM root does not exist: %s\n' "$VLLM_ROOT" >&2
|
||||||
exit 2
|
echo "[WARN] prebuilt corex check failed (non-fatal)"; return 0 2>/dev/null || true
|
||||||
}
|
}
|
||||||
[[ -f "$MANIFEST" ]] || {
|
[[ -f "$MANIFEST" ]] || {
|
||||||
printf 'prebuilt CoreX manifest is missing: %s\n' "$MANIFEST" >&2
|
printf 'prebuilt CoreX manifest is missing: %s\n' "$MANIFEST" >&2
|
||||||
exit 2
|
echo "[WARN] prebuilt corex check failed (non-fatal)"; return 0 2>/dev/null || true
|
||||||
}
|
}
|
||||||
|
|
||||||
mapfile -t artifacts < <(awk '{print $2}' "$MANIFEST")
|
mapfile -t artifacts < <(awk '{print $2}' "$MANIFEST")
|
||||||
[[ "${#artifacts[@]}" -eq 12 ]] || {
|
[[ "${#artifacts[@]}" -eq 12 ]] || {
|
||||||
printf 'expected 12 prebuilt CoreX artifacts, found %s\n' \
|
printf 'expected 12 prebuilt CoreX artifacts, found %s\n' \
|
||||||
"${#artifacts[@]}" >&2
|
"${#artifacts[@]}" >&2
|
||||||
exit 2
|
echo "[WARN] prebuilt corex check failed (non-fatal)"; return 0 2>/dev/null || true
|
||||||
}
|
}
|
||||||
|
|
||||||
for artifact in "${artifacts[@]}"; do
|
for artifact in "${artifacts[@]}"; do
|
||||||
[[ "$artifact" == corex_*.so && "$artifact" != */* ]] || {
|
[[ "$artifact" == corex_*.so && "$artifact" != */* ]] || {
|
||||||
printf 'invalid prebuilt artifact name: %s\n' "$artifact" >&2
|
printf 'invalid prebuilt artifact name: %s\n' "$artifact" >&2
|
||||||
exit 2
|
echo "[WARN] prebuilt corex check failed (non-fatal)"; return 0 2>/dev/null || true
|
||||||
}
|
}
|
||||||
done
|
done
|
||||||
|
|
||||||
|
|||||||
@@ -205,7 +205,7 @@ build_stage "skipping vLLM core block overrides (vendor_overrides not found)"
|
|||||||
fi
|
fi
|
||||||
|
|
||||||
build_stage "installing hash-pinned CoreX 3.2.3 extensions"
|
build_stage "installing hash-pinned CoreX 3.2.3 extensions"
|
||||||
bash ./install_prebuilt_corex.sh "${VLLM_ROOT}"
|
bash ./install_prebuilt_corex.sh "${VLLM_ROOT}" || echo "[WARN] install_prebuilt_corex failed (non-fatal)"
|
||||||
|
|
||||||
build_stage "compiling moe_topk_softmax CUDA kernel"
|
build_stage "compiling moe_topk_softmax CUDA kernel"
|
||||||
cd /workspace && bash ex_engine/build_moe_topk.sh 2>&1 || echo "[WARN] moe_topk build failed (non-fatal)"
|
cd /workspace && bash ex_engine/build_moe_topk.sh 2>&1 || echo "[WARN] moe_topk build failed (non-fatal)"
|
||||||
|
|||||||
Reference in New Issue
Block a user