超过128k是如何做截断的？

[llllllaz]

请问相关同学，对于上下文长度为128k的模型，应该采用哪种逻辑做截断。

[hsiehjackson]

What do you mean truncate? We directly generate data to the length you want instead of doing any truncation.

[llllllaz]

such like this