[Kernel] Integrate CUTLASS MoE kernel with PPLX (#18762)

Signed-off-by: ElizaWszola <ewszola@redhat.com>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

vllm/model_executor/layers/fused_moe/modular_kernel.py

@@ -175,6 +175,7 @@ class FusedMoEPermuteExpertsUnpermute(ABC):
     def workspace_shapes(
         self,
         a: torch.Tensor,
+        aq: torch.Tensor,
         M: int,
         N: int,
         K: int,
@@ -309,7 +310,7 @@ class FusedMoEModularKernel(torch.nn.Module):
 
         # Use a1 here to decipher the correct workspace datatype
         workspace13_shape, workspace2_shape, workspace_dtype = (
-            self.fused_experts.workspace_shapes(a1, M, N, K, top_k,
+            self.fused_experts.workspace_shapes(a1, a1q, M, N, K, top_k,
                                                 global_num_experts))
 
         # We can reuse the memory between cache1 and cache3 because by the time