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 9, 10:00 UTC
[BUGFIX][XPU] fix memory check after XPU reuse GPU_worker (#33358)
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
Chendi.Xue committed
Jan 29, 2026 at 11:56 UTC
8c8ebeb941d7bcc380b4aa685fd5201100c9eeca
1 file changed
+1
-1
vllm/v1/worker/gpu_worker.py
+1
-1
@@ -330,7 +330,7 @@ class Worker(WorkerBase):
330
free_gpu_memory = profile_result.after_profile.free_memory
331
# NOTE(woosuk): Here we assume that the other processes using the same
332
# GPU did not change their memory usage during the profiling.
333
- assert self.init_snapshot.free_memory > free_gpu_memory, (
333
+ assert self.init_snapshot.free_memory >= free_gpu_memory, (
334
"Error in memory profiling. "
335
f"Initial free memory {format_gib(self.init_snapshot.free_memory)} GiB, "
336
f"current free memory {format_gib(free_gpu_memory)} GiB. "