Training without scorer

if the training has failed to generate the scorer (this has happened because of an issue with kenlm), is the performance of recognition affected?

It is obvious that the scorer can be generate separately from the DS training, but I am not sure if the training procedure finishes without a scorer the accuracy is affected or not.

Thanks in advance,

As documented, the scorer is only used during the test phase of the training. So only WER / CER computation will be affected.

Meaning you can fix your scorer and re-run test set after with it.