positive666 / yolo_research

based on yolo-high-level project (detect\pose\classify\segment\):include yolov5\yolov7\yolov8\ core ,improvement research ,SwintransformV2 and Attention Series. training skills, business customization, engineering deployment C
GNU General Public License v3.0
756 stars 145 forks source link

GAM训练问题 #69

Closed xing-bing closed 2 years ago

xing-bing commented 2 years ago

❔Question

你好,请问一下我在网络中加入GAM注意力机制之后使用yolos可以正常训练,但是要是使用大一点的模型比如yolom的话训练时就会出现Nan

Additional context

positive666 commented 2 years ago

❔Question

你好,请问一下我在网络中加入GAM注意力机制之后使用yolos可以正常训练,但是要是使用大一点的模型比如yolom的话训练时就会出现Nan

Additional context

你是自己加的配置的YAML吗?

positive666 commented 2 years ago

❔Question

你好,请问一下我在网络中加入GAM注意力机制之后使用yolos可以正常训练,但是要是使用大一点的模型比如yolom的话训练时就会出现Nan

Additional context

这个不是模型的问题 是我合并V7的数据增强的问题 已经修复了 有问题随时联系

positive666 commented 2 years ago

❔Question

你好,请问一下我在网络中加入GAM注意力机制之后使用yolos可以正常训练,但是要是使用大一点的模型比如yolom的话训练时就会出现Nan

Additional context

需要调整下超参

github-actions[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.