You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
(Windows only) Pow operator gives incorrect result in UT test_binary_ufuncs_xpu.py::TestBinaryUfuncsXPU::test_pow_xpu_float16. #798 is to fix this failure. However, additional overhead was introduced in type cast so we need to record this.
The failure is related to the cast of complex half type in kernel. If we convert with opmath_t{}, other ops like log will also gives incorrect results.
🐛 Describe the bug
(Windows only) Pow operator gives incorrect result in UT test_binary_ufuncs_xpu.py::TestBinaryUfuncsXPU::test_pow_xpu_float16.
#798 is to fix this failure. However, additional overhead was introduced in type cast so we need to record this.
The failure is related to the cast of complex half type in kernel. If we convert with opmath_t{}, other ops like log will also gives incorrect results.
Versions
torch-xpu-ops: 47cbc58
pytorch: c8626a4e1f0553c8a9af32515555409d151a00c0
The text was updated successfully, but these errors were encountered: