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)