An Error when FT Llama2 about self-rag

[ShawnChang-ei]

oh, i solved the error by adding a line of code.
if len(tokenizer) > embedding_size:
model.resize_token_embeddings(len(tokenizer))
(+)model.config.vocab_size = len(tokenizer)

[AkariAsai]

Thank you so much for reporting the issue! Hm I though the resize_token_embeddings function will automatically update the model config but I might be wrong. Glad you found out the fix!