ImprintLab
3D训练时loss的梯度更新和参数更新
您好，
想问下，目前代码中3d训练部分，是整个video计算平均loss然后梯度回传更新参数。是否可以每一帧计算loss进行梯度回传并更新参数，这样是否就可以把video的length从2变为整个video的长度？ 涉及到memory相关的代码，应该如何修改呢？
期待并感谢您的回复！
Hello, I would like to ask, in the current 3D training part of the code, the average loss is calculated for the entire video and then the gradient is returned to update the parameters. Can the loss be calculated for each frame, the gradient is returned and the parameters are updated, so that the length of the video can be changed from 2 to the length of the entire video? How should the code related to memory be modified? Looking forward to and thank you for your reply!
It was automatically translated, probably the thread could be helpful to other users.
