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, 23:45 UTC
[Bugfix][Logprobs] Fix logprobs op to support more backend (#21591)
Signed-off-by: MengqingCao <cmq0113@163.com>
Mengqing Cao committed
Jul 25, 2025 at 20:53 UTC
f3a683b7c9df8b251092e48e53d58220bb920f2c
1 file changed
+3
-1
vllm/v1/sample/ops/logprobs.py
+3
-1
@@ -4,8 +4,10 @@
4
5
import torch
6
7
+from vllm.platforms import current_platform
8
8
-@torch.compile(dynamic=True)
9
+
10
+@torch.compile(dynamic=True, backend=current_platform.simple_compile_backend)
11
def batched_count_greater_than(x: torch.Tensor,
12
values: torch.Tensor) -> torch.Tensor:
13
"""