New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[ROCm] Skipping subtests that check support for float64 type in the NN ops #30500
[ROCm] Skipping subtests that check support for float64 type in the NN ops #30500
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@deven-amd minor cosmetic changes are required.
c8ee4e7
to
d315f3b
Compare
d315f3b
to
7276d27
Compare
I looked at the CI failure logs, and they do not seem related to the changes in this PR. The one failure in |
@deven-amd here is the internal error : |
I saw the error in the invocation log, but cannot see how the change in this PR can cause it. For the CUDA run, the only thing that will change with this PR is the ordering the dtypes in the list, it will go from Let me push out a change that keeps the above mentioned ordering intact, and see it that fixes the error! deven |
…N ops ROCm platform currently does not support the float64/double type in the NN ops This commit skips subtests (within python unit-tests) that test this functionality. The "skip" is guarded by the call to "is_built_with_rocm()", and hence these unit-tests will not be affected in any way when running with TF which was not built with ROCm support (i.e. `--config=rocm`)
7276d27
to
ca79b8d
Compare
@whchung , please re-approve this to kick-off the CI runs. Thanks |
@rthadur , the CI errors are gone :) |
@deven-amd thank you , @chsigg can you please review this again. |
…kip_double_dtyp_subtests PiperOrigin-RevId: 258814196
ROCm platform currently does not support the float64/double type in the NN ops
This commit skips subtests (within python unit-tests) that test this functionality. The "skip" is guarded by the call to "is_built_with_rocm()", and hence these unit-tests will not be affected in any way when running with TF which was not built with ROCm support (i.e.
--config=rocm
)@tatianashp @whchung @chsigg