🐛 Bug
Master branch unit test TestDynamicShape.test_nonzero_cast is failing. See CI result. Failure is:
ERROR: test_nonzero_cast (__main__.TestDynamicShape)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/tmp/pytorch/xla/test/test_operations.py", line 757, in test_nonzero_cast
t2 = torch.nonzero(t1.int()).float()
RuntimeError: /tmp/pytorch/xla/torch_xla/csrc/aten_xla_type.cpp:1132: SymIntArrayRef expected to contain only concrete integers
Per the error message, the error really looks the one I got when I enabled functionalization.
@JackCaoG , do you know how to check if functionalization is enabled?
To Reproduce
XLA_EXPERIMENTAL="nonzero:masked_select:masked_scatter" python pytorch/xla/test/test_operations.py TestDynamicShape.test_nonzero_cast
Expected behavior
The test should succeed.
Environment
- Reproducible on XLA backend [CPU/TPU]: xla::CPU
- torch_xla version: nightly.
Additional context
🐛 Bug
Master branch unit test TestDynamicShape.test_nonzero_cast is failing. See CI result. Failure is:
Per the error message, the error really looks the one I got when I enabled functionalization.
@JackCaoG , do you know how to check if functionalization is enabled?
To Reproduce
XLA_EXPERIMENTAL="nonzero:masked_select:masked_scatter" python pytorch/xla/test/test_operations.py TestDynamicShape.test_nonzero_cast
Expected behavior
The test should succeed.
Environment
Additional context