Update semi_structured_sparse.py

On my environment with an A100, I only see a speedup when enabling the following line

SparseSemiStructuredTensor._FORCE_CUTLASS = True

Author: catswe

advanced_source/semi_structured_sparse.py

@@ -55,6 +55,9 @@
 from torch.sparse import to_sparse_semi_structured, SparseSemiStructuredTensor
 from torch.utils.benchmark import Timer
 
+# the following line may need to be enabled to see a speedup
+# SparseSemiStructuredTensor._FORCE_CUTLASS = True
+
 # mask Linear weight to be 2:4 sparse
 mask = torch.Tensor([0, 0, 1, 1]).tile((3072, 2560)).cuda().bool()
 linear = torch.nn.Linear(10240, 3072).half().cuda().eval()