pere

/

nb-nn-translation

like 5

@pere Hi, very useful model thank you! However I think there is an issue with the spiece.model file. AutoTokenizer.from_pretrained("pere/nb-nn-translation", use_fast = False) gives a tokenizer with vocab size 250,100 whereas AutoTokenizer.from_pretrained("pere/nb-nn-translation", use_fast = True) gives a vocab size of 50,003. I believe the latter, which doesn't rely on the sentencepiece model directly, is the correct one. Would you be able to upload the right file? Thanks!