[webnn] Add float32 tests for WebNN hardSwish op #38712
Conversation
| 9.11885929107666, | ||
| 3.545238494873047, | ||
| 2.4567370414733887, | ||
| 0, |
There was a problem hiding this comment.
Fail to run these hardSwish tests by Chromium integrated WebNN API with XNNPack API back, since XNNPack API can distinguish 0.0 and -0.0.
@fdwr @huningxin Any suggestion? Thanks.
There was a problem hiding this comment.
Signed and unsigned zero can be treated as equivalent (IEEE does https://en.wikipedia.org/wiki/Signed_zero#Comparisons). I recommend that before checking for ULP differences in assert_array_approx_equals_ulp, first check if expected == actual, and only if !=, would you need to check ULP too.
There was a problem hiding this comment.
@fdwr Thanks for your suggestion! Yes, if actual was as same as expect one, no need to measure the ULP distance which would save testing time. Please take another look, thanks.
BruceDai
force-pushed
the
add_webnn_hardSwish_tests
branch
from
e6c322b
to
796bb20
Compare
whether actual result is as same as expect one.
BruceDai
force-pushed
the
add_webnn_hardSwish_tests
branch
from
796bb20
to
d29a6df
Compare
* [webnn] Add float32 tests for WebNN hardSwish op
* [webnn] Optimize assert_array_approx_equals_ulp method by firstly check
whether actual result is as same as expect one.
@fdwr @Honry @dontcallmedom PTAL, thanks.