lostkevin comments

Repositories
Issues
Comments

Results 1 comments of


                                            lostkevin

[BUG] Dual meaning of `max_position_embeddings`, computing both embedding shape & yarn scaling base

Hi @BoxiangW , I also observed this overwritten issue due to the duplicated name of `max_position_embeddings` and `original_max_position_embeddings` (named `max_position_embeddings` in MLATransformerConfig).