pytorch

be137e45 - reorganizing tests so that test1 and test2 are balanced in timing (#45778)

reorganizing tests so that test1 and test2 are balanced in timing (#45778) Summary: used --shard option to split up python tests ran from `test/run_test.py` in the testing script run in CI also revised a help message to be more accurate for --shard. Test results: BEFORE: | EVENT | TIMING | |---|---| | **TEST1** | | | | | | test_python_nn | 35m19s | | test_cpp_extensions | 30s | | **total** | **35m49s** | | **TEST2** | | | | | | install_torchvision | 35s | | test_python_all_except_nn_and_cpp_extensions | 255m37s | | test_aten | SKIPPED | | test_libtorch | 9m8s | | test_custom_script_ops | SKIPPED | | test_custom_backend | SKIPPED | | test_torch_function_benchmark | 10s | | **total** | **4hr24m** | AFTER THIS SHARD: | EVENT | TIMING | |---|---| | **TEST1** | | | | | | test_autograd | 26m30s | | test_foreach | 69m | | test_nn | test_nn is 35m38s | | **total** | **3h1m** | | **TEST2** | | | | | | test-quantization | 41m28s | | test_spectral_ops | 17m37s | | test_torch | 8m56s | | test_jit_legacy | 16m21s | | **total** | **2h18m** | Pull Request resolved: https://github.com/pytorch/pytorch/pull/45778 Reviewed By: albanD Differential Revision: D24137156 Pulled By: janeyx99 fbshipit-source-id: 5873fec47aedb9f699ebbda653a4d32a9950fc13