反馈 / #1613

#1613 Feature Request: Token repetition penalty

open · @comet13579 · 1 评论 · 去 GitHub 看

Models & quants

说明

When using strata with Hermes' agent, usually there will be some repetition issues, especially in IQ2_XS model and sometimes on IQ3_XXS model. Please implement that to improve the usability of lower quant models

本站相关内容

相关页面的快捷入口。