Flaky CI: worker 'gw0' crashed while running xxx

[BowenBao]

Some example runs:
https://github.com/microsoft/onnxscript/actions/runs/5138839927/jobs/9248599814?pr=755
https://github.com/microsoft/onnxscript/actions/runs/5149591642/jobs/9272793720?pr=759
https://github.com/microsoft/onnxscript/actions/runs/5159201503/jobs/9293826467?pr=758

[justinchuby]

Some flaky ort again:

https://github.com/microsoft/onnxscript/actions/runs/5225635296/jobs/9435363354#step:5:14132

[gw1] win32 -- Python 3.10.11 D:\a\onnxscript\onnxscript\.nox\test_torch_nightly\Scripts\python.EXE worker 'gw1' crashed while running 'onnxscript/tests/function_libs/torch_lib/ops_test.py::TestOutputConsistencyFullGraphCPU::test_output_match_opinfo__min_other_binary_cpu_float16'

[justinchuby]

So far they look like ops related to ReduceMin / ReduceMax; clamp_max also fails which implies Min

[BowenBao]

And so far windows only?

[justinchuby]

I've seen linux once or twice

[jcwchen]

Are these 2 tests (test_output_match_opinfo__maximum_cpu_float16 and test_output_match_opinfo__minimum_cpu) also flaky? I saw them failed on Windows in a few PRs which seem irrelevant. Since it happened quite often, perhaps we can skip these tests on Windows for now?

[justinchuby]

Sure, I will add a skip

[justinchuby]

microsoft/onnxruntime#16492