the max_position_embeddings for this model is 4096 which is way shorter than 32k of qwen2.5 models. is this intended? if so why?
· Sign up or log in to comment
