How to compute gradients of 1D-Convolution with respect to input in tensorflow?

Question

I need to compute gradients of 1D Convolution with respect to input in tensorflow. I know if this would have been 2D convolution, I could have used tf.compat.v1.nn.conv2d_backprop_input. I want to do this on 1D convolution. Is there a way to do it? Thanks.

I know about normal gradients calculation but that is not the same. What is the right way to solve this?

Answer 1

You could use GradientTape. This should work for 1D data as well es for 2D images.