旭哥在徽京
旭哥在徽京
1、update your torch to 1.6+ 2、return to use apex : "from apex import amp" and modify the code related.
i will have a try
using the newest projects is fine, model (2b-pt)
The precision of 2b-it is nice, while the 2b-pt seems like a shit, so why?
The rootcase is OOM, :100:
I do this to solve: 1. target_link_libraries( gemma stdc++fs ) 2. or update clang-version to clang-10
Debuged
可能你需要 该配置文件的 [net] [yolo] 里的配置,以兼容你的数据集和目标特性; YOLOv8本周我会使用+部署;YOLOv5-v4 v6版本我都用过,没问题的
模型没问题,你那前面正常收敛 后面发散,多半是学习率问题、降lr,或者第一个步数整体下调;不推荐使用darknet训练,和torch比掉点不少,简单部署还行