Like the tensorflow-directml build, having a nightly run scheduled regardless of whether changes were made during that day can help catch regressions due to driver updates, or catch sporadic failures. Also, some machines can be offline during a specific run and will need to be included in next day's run, even if no changes were made.
Like the
tensorflow-directml
build, having a nightly run scheduled regardless of whether changes were made during that day can help catch regressions due to driver updates, or catch sporadic failures. Also, some machines can be offline during a specific run and will need to be included in next day's run, even if no changes were made.