Yet another tacotron 2 implementation always want to implement that arch, especially on polish language, so Datasets Mozilla Common Voice pl LJSpeech dataset Architecture Pytorch implementation https://arxiv.org/abs/1712.05884