* Encodec model. * Fixes. * Add the padding functions. * Get the LSTM bit to work. * Get the encodec model to generate some tokens (decoder only for now). * Minor tweak. * Minor tweak.