How does it makes sense to apply fft on features?

I am making it super simplified, I wonder if it would make sense to apply fft to the features of a 1D signal.
Lets say I have a batch of inputs and the extracted features looks like this x.shape = BxCxL=10x64x100.
I was wondering if it makes sense if I apply fft to the features and I was wondering what the fft features would represent in that case.
I am confused what dimension I should apply the fft and how should apply it in this scenario.
Any suggestion would be really appreciated!

this may help