06-14, 14:00–14:35 (Europe/Prague), E105 (capacity 70)
Traditionally, most test failure analysis tools have relied on text output but OpenQA tests are primarily visual in nature which makes it difficult to use those more common techniques. In Fedora, we have been exploring the use of AI/ML to do some test failure classification using the primarily visual information that we have available. This talk will cover the techniques that we currently evaluating and using for test failure classification for OpenQA test failures in Fedora.
Tim has been working as part of Fedora QA since 2011 and has been focusing on AI/ML, GPU acceleration in Fedora and using AI/ML in testing as of late. Tim earned his MSCS degree focusing in machine learning and computational lingustics in 2023.