Tests

Merge pull request #109 from pytorch/_C-instead-of-load_library #34

Job	Run time
unittests-linux (3.13, linux.g5.4xlarge.nvidia.gpu, cuda, 12.4) / linux-job	7m 12s
unittests-linux (3.11, linux.12xlarge, cpu) / linux-job	3m 5s
	10m 17s

Provide feedback