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 15, 16:00 UTC
[V1][Bugfix] Fix typo in MoE TPU checking (#15927)
Signed-off-by: Roger Wang <ywang@roblox.com>
Roger Wang committed
Apr 1, 2025 at 23:46 UTC
0e00d40e4faa1c433d9e16fbca82ba961c80c649
1 file changed
+1
-1
vllm/model_executor/layers/fused_moe/layer.py
+1
-1
@@ -309,7 +309,7 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
309
expert_map=expert_map,
310
renormalize=renormalize)
311
312
- forward_native = forward_tpu if current_platform.is_tpu else forward_cuda
312
+ forward_native = forward_tpu if current_platform.is_tpu() else forward_cuda
313
314
315
def determine_expert_map(