How can I call adaptive average pooling function for 4D tensor

[ifmaq1]

1-) I have a .pth image features' file with 4D tensor [batch,dimension,height,width]. I can see this class (AdaptiveAveragePooling2D(AdaptivePooling2D):

Could you please tell me how can I Pass my .pth file tensor to this class in order to get the pooled features as output?

2-) How to use this function

tfa.layers.AdaptiveAveragePooling2D( output_size: Union[int, Iterable[int]], data_format=None, kwargs )

3-) I want final tensor of shape [2048,36]