Why does the same code train much slower in pytorch than tensorflow

[ZZZ77QQQ]

Hello, I would like to ask, I used the same code to run under the framework of tensorflow and pytorch respectively, and found that the training speed in pytorch is much slower than that in tensorflow, what is the reason?

[lululxvi]

Pytorch in general is slow.

[ZZZ77QQQ]

Thank you for your reply, I would like to ask a question, that is, when doing the inverse problem, if the parameter that you want to solve is not a fixed parameter but a variable parameter, is there a way to solve it

[yuluoxiansen]

Hello, for the sampling points in deepxde, can I take different numbers of points in area a and area b respectively? I want to take more points in a to better capture the characteristics of pde.

[lululxvi]

@ZZZ77QQQ Yes, there are demo examples.

[ZZZ77QQQ]

Thank you for your reply, I have another question to ask. In DeepONet, if the input of the trunk layer is a time series, is it necessary to normalize the time series?

[lululxvi]

Could be helpful.