-
Notifications
You must be signed in to change notification settings - Fork 245
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
训练几个epoch召回率和loss都变成0 #37
Comments
100个epoch之后再看 |
好的~ |
方便的话,能否提供下您的预训练模型? |
目前没有预训练 |
好的,仍然感谢您! |
那我先关了这个了 |
@VixeruntR 您这边往后训练多个epoch后是正常的吗,我训练也是,第三轮开始recall就是0了,看着有点慌 |
我刚开始数据集大概2000张图像出现了这种情况,后来加了大概4000的icdar里面的数据,学习率设置的5e-4,batch_size=8,主干是resnet18,就没出现这个情况了,训练了20个epoch,recall 大概0.6,实际效果就已经非常好了。希望对你有帮助 |
感谢!我在resnet18下面训收敛比较快,换成resnet50就很慢了,100个epoch后也不见多好,效果还比上训练十几轮的resnet18 |
我也是。。训练20轮的resnet18是最香的。。 |
训练几百张数据是不是容易过拟合啊 |
几百张有点太少了。。 |
@VixeruntR 你好,请问,我用了几百张图片训练,1200轮都跑完了,P,R,F仍然是0,是什么原因呢?几百张可能效果不好,不不至于全都是0啊 |
有可能就是你的pyrorch版本跟cuda版本不匹配 |
第一个epoch召回率0.5左右,然后一直变低,同时loss也在变低,大概四五个epoch之后突然全部变0,这可能是哪方面的原因呢?
The text was updated successfully, but these errors were encountered: