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 14, 05:30 UTC
[Misc][gpt-oss] guard import when triton kernel when not up to date (#22584)
Signed-off-by: zhewenli <zhewenli@meta.com>
Zhewen Li committed
Aug 10, 2025 at 21:29 UTC
afa5b7ca0b417abadfa85e32f28969b72e58a885
1 file changed
+12
-3
vllm/model_executor/layers/fused_moe/gpt_oss_triton_kernels_moe.py
+12
-3
@@ -5,15 +5,24 @@ from typing import TYPE_CHECKING, Any, Optional
5
import torch
6
7
import vllm.model_executor.layers.fused_moe.modular_kernel as mk
8
+from vllm.logger import init_logger
9
from vllm.model_executor.layers.fused_moe.topk_weight_and_reduce import (
10
TopKWeightAndReduceDelegate)
11
from vllm.model_executor.layers.fused_moe.utils import extract_required_args
12
from vllm.utils import has_triton_kernels
13
14
+logger = init_logger(__name__)
15
+
16
if has_triton_kernels():
14
- import triton_kernels.swiglu
15
- from triton_kernels.matmul_ogs import FnSpecs, FusedActivation, matmul_ogs
16
- from triton_kernels.routing import routing
17
+ try:
18
+ import triton_kernels.swiglu
19
+ from triton_kernels.matmul_ogs import (FnSpecs, FusedActivation,
20
+ matmul_ogs)
21
+ from triton_kernels.routing import routing
22
+ except ModuleNotFoundError:
23
+ logger.error(
24
+ "Failed to import Triton kernels. Please make sure your triton "
25
+ "version is compatible.")
26
27
if TYPE_CHECKING:
28
from triton_kernels.matmul_ogs import PrecisionConfig