Skip to content
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.

Sign up for GitHub

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

[TF][GPU]: Enable bitwise shift test in TF. #26194

Merged
merged 9 commits into from
Sep 22, 2024
Merged

[TF][GPU]: Enable bitwise shift test in TF. #26194

Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
6d7fb07
[TF][GPU]: Enable bitwise shift test in TF.
pkowalc1 Aug 23, 2024
aecb276
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
rkazants Aug 23, 2024
894fd44
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
rkazants Aug 23, 2024
64ed042
[TF]: Enabled bitwise shift tests for datatype supported on GPU
pkowalc1 Aug 26, 2024
b9f2c2b
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
pkowalc1 Aug 29, 2024
8b2bd4a
[TF]:[BitwiseShift]: Disabled test for not working on GPU type.
pkowalc1 Aug 29, 2024
f6d9e93
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
pkowalc1 Aug 30, 2024
d0bc689
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
pkowalc1 Aug 30, 2024
e013b82
Merge branch 'master' into bitwise_shift_gpu_enable_tf_tests
pkowalc1 Sep 20, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions tests/layer_tests/tensorflow_tests/test_tf_BitwiseShift.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,8 +68,8 @@ def create_bitwise_net(self, x_shape, y_shape, is_y_const, input_type, op_type):
@pytest.mark.nightly
def test_bitwise(self, x_shape, y_shape, is_y_const, input_type, op_type, ie_device, precision, ir_version,
temp_dir, use_legacy_frontend):
if ie_device == 'GPU':
pytest.skip("149424: Bitwise ops are not supported on GPU")
if ie_device == 'GPU' and input_type in [np.uint64]:
pytest.skip("149424: uint64 type is not supported on GPU")
Copy link
Member

@rkazants rkazants Aug 29, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please create an issue ticket for GPU plugin to support uint64. And point the right ticket number here.

Copy link
Member

@rkazants rkazants Sep 3, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@pkowalc1, please create separate ticket for uint64 support on GPU and list this ticket number in the skipping code

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@rkazants if you think there is a real need for uint64 support on GPU, please create a ticket and motivate it properly(e.g. maybe there is a business need). Currently, uint64 on GPU will be converted to int32 (info from V. Paramuzow), so there is a support but it it will not work properly in this case.

Copy link
Member

@rkazants rkazants Sep 4, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The motivation is to be fully aligned with framework cases and capabilities. Our intention is to support not only known models but all other models about which we don't know.
Please create ticket and list here if you are responsible for resolving this task and if you want to change TF FE test sources. Otherwise, I can switch on these tests on GPU in my separate PR and create ticket on my own.
@mlukasze, fyi, I cannot allow to merge this PR without JIRA ticket about this limitation for uint64_t.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@rkazants This PR is ready to be merged - all types except u64(as discussed above) should be working fine. However, if you want to create your own PR - no problem, please just close this one.

if use_legacy_frontend:
pytest.skip("BitwiseShift ops are supported only by new TF FE")
self._test(*self.create_bitwise_net(x_shape=x_shape, y_shape=y_shape, is_y_const=is_y_const,
Expand Down
Loading