Training Mozilla TTS on a K8s/Spark Cluster

Good day all…

Taking a first shot at training a LibriTTS model using tacotron on a K8s/Spark cluster. I’m interested in if anyone else has been successful setting something like this up? Any advice?

Is this approach even feasible / recommended? I’m trying to address training larger datasets in reasonable timeframes.

Any thoughts / suggestions / cautions would be appreciated!