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, 04:15 UTC
[Misc] add error message in non linux platform (#3438)
youkaichao committed
Mar 15, 2024 at 14:21 UTC
604f235937684aa173afec96cca5b02e3b0bc154
1 file changed
+8
-1
setup.py
+8
-1
@@ -9,12 +9,17 @@ from typing import List, Set
9
10
from packaging.version import parse, Version
11
import setuptools
12
+import sys
13
import torch
14
import torch.utils.cpp_extension as torch_cpp_ext
15
from torch.utils.cpp_extension import BuildExtension, CUDAExtension, CUDA_HOME, ROCM_HOME
16
17
ROOT_DIR = os.path.dirname(__file__)
18
19
+# vLLM only supports Linux platform
20
+assert sys.platform.startswith(
21
+ "linux"), "vLLM only supports Linux platform (including WSL)."
22
+
23
# If you are developing the C++ backend of vLLM, consider building vLLM with
24
# `python setup.py develop` since it will give you incremental builds.
25
# The downside is that this method is deprecated, see
@@ -402,11 +407,13 @@ def get_vllm_version() -> str:
407
if neuron_version != MAIN_CUDA_VERSION:
408
neuron_version_str = neuron_version.replace(".", "")[:3]
409
version += f"+neuron{neuron_version_str}"
405
- else:
410
+ elif _is_cuda():
411
cuda_version = str(nvcc_cuda_version)
412
if cuda_version != MAIN_CUDA_VERSION:
413
cuda_version_str = cuda_version.replace(".", "")[:3]
414
version += f"+cu{cuda_version_str}"
415
+ else:
416
+ raise RuntimeError("Unknown runtime environment.")
417
418
return version
419