fine-tuning
#8
by imaqn - opened
does fine-tuning will work by adding <eou> token in the end of utterance for each complete sentences inside the dataset?
for example:
hello, how are you? <eou> (complete sentence, token added)
oh, I think this is (incomplete sentence, no token added)
Also, is there any fine-tuning script we can use?
imaqn changed discussion status to closed