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 14, 05:30 UTC
[BugFix] Fix topk_softmax assert (#19764)
Signed-off-by: Luka Govedic <lgovedic@redhat.com>
Luka Govedič committed
Aug 27, 2025 at 12:47 UTC
4f35be10a96feeca0328d3ab8d359e1eaae5c23d
1 file changed
+1
-1
csrc/moe/topk_softmax_kernels.cu
+1
-1
@@ -573,7 +573,7 @@ void topk_softmax(
573
stream);
574
}
575
else {
576
- assert(topk_indices.scalar_type() == at::ScalarType::Int64);
576
+ TORCH_CHECK(topk_indices.scalar_type() == at::ScalarType::Long);
577
vllm::moe::topkGatingSoftmaxKernelLauncher(
578
gating_output.data_ptr<float>(),
579
topk_weights.data_ptr<float>(),