This is a zipformer ASR+AT model. The model is trained on LibriSpeech + AudioSet (no weighted sampler)
-