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 4, 15:00 UTC
[CI][AMD]BugFix] Fix deadlock occuring in test_moe_layer (#40767)
Signed-off-by: Randall Smith <Randall.Smith@amd.com>
rasmith committed
Apr 25, 2026 at 08:34 UTC
1e9f19ca3fd29f83442ab83b08d4642e691c95bd
1 file changed
+1
tests/kernels/moe/test_moe_layer.py
+1
@@ -1632,6 +1632,7 @@ def _parallel_worker(
1632
if all2all_manager is not None:
1633
all2all_manager.destroy()
1634
total = total + 1
1635
+ torch.distributed.barrier()
1636
1637
skipped = total - (passed + failed)
1638