Hello,
Let me bring what i have faced with during testing the functionality.
Describe the bug
It has been observed that some Catch2 tests for PCA executed with bazel sporadically fail between testing runs with lack of precision on the mean/variance calculation during training stage. An example:
During investigation it has been observed that "sporadic" fails become regular with launching tests in a single-threaded mode. This fact might be helpful for validation.
To Reproduce
Steps to reproduce the behavior:
- Download
oneDAL and bazel (bazelisk-linux-amd64)
bazel build //cpp/oneapi/dal/algo/pca:test_batch_host --config=host --test_link_mode=dev --test_thread_mode=par --test_external_datasetstaskset -c 0-0 ./bazel-bin/cpp/oneapi/dal/algo/pca/test_batch_host
Expected behavior
Tests have passed successfully.
Output/Screenshots
Screen of log with tests output.
Environment:
- OS: Ubuntu 20.04
- Compiler: GCC9.4
- Bazel: 6.3.2
- Version: latest
Hello,
Let me bring what i have faced with during testing the functionality.
Describe the bug
It has been observed that some Catch2 tests for PCA executed with bazel sporadically fail between testing runs with lack of precision on the mean/variance calculation during training stage. An example:
During investigation it has been observed that "sporadic" fails become regular with launching tests in a single-threaded mode. This fact might be helpful for validation.
To Reproduce
Steps to reproduce the behavior:
oneDALandbazel(bazelisk-linux-amd64)bazel build //cpp/oneapi/dal/algo/pca:test_batch_host --config=host --test_link_mode=dev --test_thread_mode=par --test_external_datasetstaskset -c 0-0 ./bazel-bin/cpp/oneapi/dal/algo/pca/test_batch_hostExpected behavior
Tests have passed successfully.
Output/Screenshots
Screen of log with tests output.
Environment: