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 5, 07:15 UTC
[XPU] Disable CUDA graph memory estimate on XPU platform (#41344)
Signed-off-by: chaojun-zhang <chaojun.zhang@intel.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
Chaojun Zhang committed
May 6, 2026 at 16:38 UTC
1c58876618dc7275520297cbbd1244a64f551952
1 file changed
+1
-1
vllm/v1/worker/gpu_worker.py
+1
-1
@@ -398,7 +398,7 @@ class Worker(WorkerBase):
398
# differently and can produce incorrect/negative estimates.
399
cudagraph_memory_estimate = 0
400
if (
401
- not current_platform.is_rocm()
401
+ current_platform.is_cuda()
402
and self.vllm_config.compilation_config.cudagraph_mode
403
!= CUDAGraphMode.NONE
404
):