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, 09:45 UTC
[BugFix] fix tests/kernels/moe/test_moe_layer.py (#39404)
Signed-off-by: Richard Zou <zou3519@gmail.com>
Richard Zou committed
Apr 9, 2026 at 14:48 UTC
6c749399b7dc939a592e30d5964d7bdf255427aa
1 file changed
+1
-1
vllm/model_executor/layers/fused_moe/runner/moe_runner_base.py
+1
-1
@@ -40,7 +40,7 @@ from vllm.utils.torch_utils import (
40
41
def get_layer_from_name(layer_name: str) -> torch.nn.Module:
42
forward_context: ForwardContext = get_forward_context()
43
- if layer_name == "from_forward_context":
43
+ if not HAS_OPAQUE_TYPE and layer_name == "from_forward_context":
44
all_moe_layers = forward_context.all_moe_layers
45
assert all_moe_layers is not None
46
moe_layer_index = forward_context.moe_layer_index