I'm trying to run MiniLLM with Llama-2-7b as the teacher model. Not sure why this error is persisting -- my arguments passed into forward() in the PPOModel class do not contain token_type_id and I have included return_token_type_ids=False with the AutoTokenizer class.
All related issues I found online told me to upgrade the transformers version, but that doesn't seem to work because ParallelLlamaForCausalLM is implemented locally.
I'm trying to run MiniLLM with Llama-2-7b as the teacher model. Not sure why this error is persisting -- my arguments passed into forward() in the
PPOModel
class do not contain token_type_id and I have includedreturn_token_type_ids=False
with the AutoTokenizer class.All related issues I found online told me to upgrade the transformers version, but that doesn't seem to work because
ParallelLlamaForCausalLM
is implemented locally.