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, 11:00 UTC
Fix Broken macro for cutlass moe (#18049)
Signed-off-by: drisspg <drisspguessous@gmail.com>
Driss Guessous committed
May 12, 2025 at 23:31 UTC
e57e4d6e9e3aa9987c1cffe4724d59d52b97c44e
1 file changed
+2
-1
csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu
+2
-1
@@ -29,7 +29,8 @@ void cutlass_scaled_mm_sm90(torch::Tensor& c, torch::Tensor const& a,
29
torch::Tensor const& a_scales,
30
torch::Tensor const& b_scales,
31
std::optional<torch::Tensor> const& bias);
32
-
32
+#endif
33
+#if defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90
34
void cutlass_moe_mm_sm90(
35
torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
36
torch::Tensor const& b_tensors, torch::Tensor const& a_scales,