This allows us to detect test cases that store their .mlir or .mlirbc file remotely, since the test_data_flags.txt or test_cases.json file will always exist locally. This also changes test case names in summaries from e.g.
PASSED onnx/node/generated/test_sub_uint8/model.mlir::cpu_llvm_sync_test
PASSED onnx/node/generated/test_sub_example/model.mlir::cpu_llvm_sync_test
PASSED onnx/node/generated/test_sub_bcast/model.mlir::cpu_llvm_sync_test
XFAIL pytorch/models/opt-125M/opt-125M.mlirbc::cpu_llvm_task_splats - Expected compilation to fail (included in 'expected_compile_failures')
to e.g.
PASSED onnx/node/generated/test_sub_example/test_data_flags.txt::model.mlir::cpu_llvm_sync
PASSED onnx/node/generated/test_sub_uint8/test_data_flags.txt::model.mlir::cpu_llvm_sync
PASSED onnx/node/generated/test_sub_bcast/test_data_flags.txt::model.mlir::cpu_llvm_sync
XFAIL pytorch/models/opt-125M/test_cases.json::opt-125M.mlirbc::cpu_llvm_task::splats - Expected compilation to fail (included in 'expected_compile_failures'
(anchored on the .txt or .json so added back the .mlir, also dropped the "test")
This allows us to detect test cases that store their
.mliror.mlirbcfile remotely, since thetest_data_flags.txtortest_cases.jsonfile will always exist locally. This also changes test case names in summaries from e.g.to e.g.
(anchored on the .txt or .json so added back the .mlir, also dropped the "test")
This is a bit hacky either way. I took a look at https://docs.pytest.org/en/stable/example/customdirectory.html as an alternative to https://docs.pytest.org/en/stable/example/nonpython.html and that could help.