interesting, I could note the token for each character in the captions, but then I'd also have to no
interesting, I could note the token for each character in the captions, but then I'd also have to note it separately in the training parameters? Is there a risk of it bleeding?











