About train data : time lenth ; and some code parameters : such as STFT window parameters and so on

[zhoudongJenny]

Dear Professor Zhu，I'm a beginner, so I'm sorry to bother you if I ask some stupid questions； please bear with me and forgive me.
I encountered some problems when using the project：

1. The training data I use is very long, for example, 1800 seconds, the itp is 600 seconds, its value is 60000; and the DeepDenoiser original plan comes with 30s. Do I need to change a lot of parameters?
   （For example, nperseg nfft nt X_shape, are there any principles for selecting these parameters?）

2. Because my data length is long ,so do neural network parameters need to be selected? For example, the size of the convolution kernel.

The above is my question, and I sincerely look forward to your reply;
Thank you so much.
Best Regards