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
[Hardware][CPU] Disable torch.compile for RISC-V to prevent APIError (#26693)
Signed-off-by: lyd1992 <liuyudong@iscas.ac.cn> Signed-off-by: ihb2032 <1355790728@qq.com> Signed-off-by: lyd1992 <liuyudong@iscas.ac.cn
ihb2032 committed
Oct 13, 2025 at 22:56 UTC
4a61950f4d154d6518898a46ca7f1c0e2caaacdf
1 file changed
+5
-2
vllm/v1/sample/ops/topk_topp_sampler.py
+5
-2
@@ -9,7 +9,7 @@ from packaging import version
9
from vllm import envs
10
from vllm.config.model import LogprobsMode
11
from vllm.logger import init_logger
12
-from vllm.platforms import current_platform
12
+from vllm.platforms import CpuArchEnum, current_platform
13
14
logger = init_logger(__name__)
15
@@ -73,7 +73,10 @@ class TopKTopPSampler(nn.Module):
73
)
74
self.forward = self.forward_native
75
elif current_platform.is_cpu():
76
- self.forward = self.forward_cpu
76
+ if current_platform.get_cpu_architecture() == CpuArchEnum.RISCV:
77
+ self.forward = self.forward_native
78
+ else:
79
+ self.forward = self.forward_cpu
80
else:
81
self.forward = self.forward_native
82