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
[PyTorch Vulkan] fix bug of aten::cat
for concatenation of 3D tensors at channel dim with channels as multiple of 4
#103718
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/103718
Note: Links to docs will display an error until the docs builds have been completed. ✅ 4 Unrelated FailuresAs of commit ee6d0b0: BROKEN TRUNK - The following job failed but were present on the merge base 15eed5b:👉 Rebase onto the `viable/strict` branch to avoid these failures
UNSTABLE - The following jobs failed but were likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This pull request was exported from Phabricator. Differential Revision: D46755034 |
91e5c5f
to
e1bf5bf
Compare
…rs at channel dim with channels as multiple of 4 (pytorch#103718) Summary: Pull Request resolved: pytorch#103718 The original `cat_feature_mult4ch` assumes input tensors are of 4d and use `tensor.sizes()[1]` to obtain the channel info of the tensor. This will cause bugs when the input tensors are of 3D. We generalize `cat_feature_mult4ch` to make it cover both 3D and 4D. Test Plan: Test for 3D tensors with channels as multiple of 4 is show below. Full test result is in P771032677. ``` (base) luwei@luwei-mbp fbsource % buck run --target-platforms ovr_config//platform/macos:arm64-fbsource //xplat/caffe2:pt_vulkan_api_test_binAppleMac\#macosx-arm64 -c pt.vulkan_full_precision=1 -- --gtest_filter="*cat_3d_dim0_mult4ch_success*" Building: finished in 0.1 sec (100%) 263/2812 jobs, 0/2812 updated Total time: 0.1 sec BUILD SUCCEEDED Running main() from xplat/third-party/gmock/googletest-1.12.1/googletest/src/gtest_main.cc Note: Google Test filter = *cat_3d_dim0_mult4ch_success* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from VulkanAPITest [ RUN ] VulkanAPITest.cat_3d_dim0_mult4ch_success [ OK ] VulkanAPITest.cat_3d_dim0_mult4ch_success (129 ms) [----------] 1 test from VulkanAPITest (129 ms total) [----------] Global test environment tear-down [==========] 1 test from 1 test suite ran. (129 ms total) [ PASSED ] 1 test. ``` Reviewed By: SS-JIA Differential Revision: D46755034 fbshipit-source-id: c6db3f9da452343bb9582e0712982063580a67a5
This pull request was exported from Phabricator. Differential Revision: D46755034 |
…rs at channel dim with channels as multiple of 4 (pytorch#103718) Summary: Pull Request resolved: pytorch#103718 The original `cat_feature_mult4ch` assumes input tensors are of 4d and use `tensor.sizes()[1]` to obtain the channel info of the tensor. This will cause bugs when the input tensors are of 3D. We generalize `cat_feature_mult4ch` to make it cover both 3D and 4D. Test Plan: Test for 3D tensors with channels as multiple of 4 is show below. Full test result is in P771032677. ``` (base) luwei@luwei-mbp fbsource % buck run --target-platforms ovr_config//platform/macos:arm64-fbsource //xplat/caffe2:pt_vulkan_api_test_binAppleMac\#macosx-arm64 -c pt.vulkan_full_precision=1 -- --gtest_filter="*cat_3d_dim0_mult4ch_success*" Building: finished in 0.1 sec (100%) 263/2812 jobs, 0/2812 updated Total time: 0.1 sec BUILD SUCCEEDED Running main() from xplat/third-party/gmock/googletest-1.12.1/googletest/src/gtest_main.cc Note: Google Test filter = *cat_3d_dim0_mult4ch_success* [==========] Running 1 test from 1 test suite. [----------] Global test environment set-up. [----------] 1 test from VulkanAPITest [ RUN ] VulkanAPITest.cat_3d_dim0_mult4ch_success [ OK ] VulkanAPITest.cat_3d_dim0_mult4ch_success (129 ms) [----------] 1 test from VulkanAPITest (129 ms total) [----------] Global test environment tear-down [==========] 1 test from 1 test suite ran. (129 ms total) [ PASSED ] 1 test. ``` Reviewed By: SS-JIA Differential Revision: D46755034 fbshipit-source-id: 112216ff29bbe50ef7e8723782c7d0322beddc63
e1bf5bf
to
ee6d0b0
Compare
This pull request was exported from Phabricator. Differential Revision: D46755034 |
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Summary: The original
cat_feature_mult4ch
assumes input tensors are of 4d and usetensor.sizes()[1]
to obtain the channel info of the tensor. This will cause bugs when the input tensors are of 3D. We generalizecat_feature_mult4ch
to make it cover both 3D and 4D.Test Plan:
Test for 3D tensors with channels as multiple of 4 is show below. Full test result is in P771032677.
Reviewed By: SS-JIA
Differential Revision: D46755034