what is the difference between 1x1 conv and avergaepool1d ?
A 1x1
conv layer will use a kernel size of 1x1
(i.e. a single pixel) and would act as a linear layer in the default setup. The nn.AdaptiveAvgPool1d
layer works on 3D inputs in the shape [batch_size, channels, seq_len]
and returns a desired output shape in the temporal dimension by applying an adaptive windowing approach to calculate the average of the input values.
1 Like