Fix the broken gemm test

Summary: As the title goes.

Reviewed By: bertmaher

Differential Revision: D64480822

fbshipit-source-id: ec1d17be0619fb35d4d8f774eab2858e75afe2e3

Author: xuzhao9

torchbenchmark/operators/gemm/operator.py

@@ -149,10 +149,11 @@ def triton_tutorial_matmul(self, a, b, bias) -> Callable:
 
     @register_benchmark()
     def matmul_partition_k(self, a, b, bias) -> Callable:
+        bc = b.contiguous()
         if not bias == None:
-            return lambda: matmul_partition_k(a, b) + bias
+            return lambda: matmul_partition_k(a, bc) + bias
         else:
-            return lambda: matmul_partition_k(a, b)
+            return lambda: matmul_partition_k(a, bc)
 
     @register_benchmark()
     def triton_persistent_matmul(self, a, b, bias) -> Callable: