@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / ad28a74beb
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 16, 18:15 UTC

[misc][cuda] add warning for pynvml user (#7675)

youkaichao committed Aug 20, 2024 at 00:35 UTC ad28a74beb8497a6961cfb39c63d6070712f1ef7
1 file changed +6
vllm/platforms/cuda.py
+6
@@ -18,6 +18,12 @@ logger = init_logger(__name__)
18 _P = ParamSpec("_P")
19 _R = TypeVar("_R")
20
21 +if pynvml.__file__.endswith("__init__.py"):
22 + logger.warning(
23 + "You are using a deprecated `pynvml` package. Please install"
24 + " `nvidia-ml-py` instead. See https://pypi.org/project/pynvml "
25 + "for more information.")
26 +
27 # NVML utils
28 # Note that NVML is not affected by `CUDA_VISIBLE_DEVICES`,
29 # all the related functions work on real physical device ids.