I want to use the speech_commands dataset for a simple keyword recognition model. But I simply cant get my preprocessing pipeline to work.
