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 15, 04:45 UTC
[CI][CPU] Improve dummy Triton interfaces and fix the CPU CI (#19838)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Li, Jiang committed
Jun 19, 2025 at 23:46 UTC
01220ce89a332a4105c0031933c5079036ceefa0
2 files changed
+2
-3
requirements/cpu.txt
-3
@@ -21,9 +21,6 @@ torchvision; platform_machine != "ppc64le" and platform_machine != "s390x"
21
torchvision==0.22.0; platform_machine == "ppc64le"
22
datasets # for benchmark scripts
23
24
-# cpu cannot use triton 3.3.0
25
-triton==3.2.0; platform_machine == "x86_64"
26
-
24
# Intel Extension for PyTorch, only for x86_64 CPUs
25
intel-openmp==2024.2.1; platform_machine == "x86_64"
26
intel_extension_for_pytorch==2.7.0; platform_machine == "x86_64"
vllm/triton_utils/importing.py
+2
@@ -68,9 +68,11 @@ class TritonPlaceholder(types.ModuleType):
68
69
def __init__(self):
70
super().__init__("triton")
71
+ self.__version__ = "3.3.0"
72
self.jit = self._dummy_decorator("jit")
73
self.autotune = self._dummy_decorator("autotune")
74
self.heuristics = self._dummy_decorator("heuristics")
75
+ self.Config = self._dummy_decorator("Config")
76
self.language = TritonLanguagePlaceholder()
77
logger.warning_once(
78
"Triton is not installed. Using dummy decorators. "