Generating distillation data of a competitor's model could also be a valid reason to do that. You need a huge amount of tokens to do that (that you don't want to pay yourself), and you would like to use as many…
Generating distillation data of a competitor's model could also be a valid reason to do that. You need a huge amount of tokens to do that (that you don't want to pay yourself), and you would like to use as many…