How can I get the kernel Matrix values of the CNN?

Question

I am new to machine learning. I am working with a Neural Network that implements the U-Net.
The thing which I understand is when a convolutional layer is created:

createsconv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1)

let the in_channel = 3 and out_channel = 64 then,
for kernel_size = 3 it computes with a 3*3 Kernel.

But my question is what are the values of this kernel? How it gets the values?

Answer 1

Those are the learnable parameters which are learnt through the usual back-propagation algorithm. They can be accessed before and after the learning phase via nn.weight as described in the PyTorch manual.