Hi, currently I am training the speech recognition from the scratch and I following the tutorial in the website. I am using my own dataset for the training process in Train Speech Recognizer step. But I got this error, could someone help me with this error? It said size of tensor A must match the size of tensor B at non-singleton dimension 1. Thanks a lot!
Probably some datas in your dataset is so long. You can print batch.id in order to check which datas causes this error. For example, can you try with files less than 20 seconds or smaller batch size?