Training Speech Recognition on my custom language dataset

So I’m training a Vietnamese ASR based on the ASRfromScratch tutorial, and at the ASR model training step, I understand that I have to modify the pretrained path in the train.yaml to the local folder containing Tokenizer and LM checkpoint file from the previous steps (but they’re not in a same folder, in I had to add 2 different paths for lm and tokenizer at the end of yaml file, but what about the asr.ckpt? I tried keeping the defaut one (with mini_libri data) but it doesn’t work, and I get a bunch of errors like this… I don’t know where I messed it up and how to properly do this please help me

1 Like

It is most likely that something went wrong between the declaration of the RNNLM (at training stage for RNNLM) and ASR yaml. Please look at the parameters in both yaml to see if there is an error in the dimensions.