@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / c7f2cf2b7f
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 17, 01:45 UTC

[CI] Reduce wheel size by not shipping debug symbols (#4602)

Simon Mo committed May 4, 2024 at 21:28 UTC c7f2cf2b7f67bce5842fedfdba508440fe257375
2 files changed +5
.buildkite/check-wheel-size.py
+3
@@ -25,6 +25,9 @@ def check_wheel_size(directory):
25 f"compare to the allowed size ({MAX_SIZE_MB} MB).")
26 print_top_10_largest_files(wheel_path)
27 return 1
28 + else:
29 + print(f"Wheel {wheel_path} is within the allowed size "
30 + f"({wheel_size_mb} MB).")
31 return 0
32
33
.github/workflows/publish.yml
+2
@@ -79,6 +79,8 @@ jobs:
79
80 - name: Build wheel
81 shell: bash
82 + env:
83 + CMAKE_BUILD_TYPE: Release # do not compile with debug symbol to reduce wheel size
84 run: |
85 bash -x .github/workflows/scripts/build.sh ${{ matrix.python-version }} ${{ matrix.cuda-version }}
86 wheel_name=$(ls dist/*whl | xargs -n 1 basename)