FloatingPointError: Predicted boxes or scores contain Inf/Nan. Training has diverged.

深度学习训练PointRend网络时,configs文件是"PointRend/configs/InstanceSegmentation/pointrend_rcnn_X_101_32x8d_FPN_3x_coco.yaml",因为两个人同时使用服务器GPU的原因,我把batch_size设置为1才能开始训练,但迭代600多次就抛出错误:

Error:FloatingPointError: Predicted boxes or scores contain Inf/Nan. Training has diverged.

    经查阅,是learning_raye设置太大的原因,当时我的学习率是0.02,后来改成0.001就可以完整训练了。参考博客:https://ask.csdn.net/questions/7665290 

你可能感兴趣的:(深度学习,pytorch)