AutoEncoder Inference for Version 1
Opened this issue · 0 comments
Haoliu-cola commented
Hi, thanks for your great work. I would like to test the autoencoding part in your v1 model. I tried to adjust line 89 in ft_inference.py to prompt_ids = [model.ae_token_id], but got empty generated outcomes. Could you advise on how to get a correct AE output? Thanks!