metadata
task_categories:
- text-to-audio
This dataset is used to train JET2001/speecht5_tts_imda_nsc_p1_p3_v2
It is composed of 90% of training data from parts P1 and P3, 5% of validation only from Part 3, and 5% of testing data only from Part 3.
The input_ids are processed from SpeechT5Processor.tokenizer