UNQ C3 issues - not sure if right sizes are used unpacking feature maps

I filled in UNQ C3 as described in the instructions with appropriate context hints and didn’t get appropriate output shown here:


for the second instruction, can you describe if the intended direction is just to unpack the image feature map directly?

Hi @Bradley_M_Messer. This set of unit tests is checking to make sure that the noise you are injecting is the same for each channel. That’s what these instructions are trying to explain:

The noise tensor is not entirely random; it is initialized as one random channel that is then multiplied by learned weights for each channel in the image. For example, imagine an image has 512 channels and its height and width are (4 x 4). You would first create a random (4 x 4) noise matrix with one channel.

Check how you defined noise_shape and make sure it is only defined for one channel. This will then be multiplied by the self.weight which will hold the learned values for each channel.