Hi,
I face an issue with my wav2vec model, that words tends to merged. i.e. the space between words not recognized. (“thespace between wordsnot recognized”).
It’s more common when the speaker talk relativly fast, but not only.
Any idea how to solve it?