as I have read here, these lines codes of maxpooling
- out_height = util.simplify((height - kernel_height + pad_top + pad_down) // stride_height + 1)
- out_width = util.simplify((width - kernel_width + pad_left + pad_right) // stride_width + 1)
- dheight = tvm.reduce_axis((0, kernel_height))
- dwidth = tvm.reduce_axis((0, kernel_width))
- if pool_type == ‘max’:
-
temp = pad(data, pad_before, pad_after, name="pad_temp", \ -
pad_value=tvm.min_value(data.dtype)) -
return tvm.compute((batch, channel, out_height, out_width), \ -
lambda n, c, h, w: \ -
tvm.max(temp[n, c, h*stride_height+dheight, w*stride_width+dwidth], \ -
axis=[dheight, dwidth]), \ -
tag="pool_max")
in tvm.max function, why should be " h * stride_height + dheight " here, but not " h * kernel_height " ?
for my understanding, h is belong to [0, out_height], " h * kernel_height " is the expend height of matrix that need to be done max pooling. the same for " w*stride_width " …
thanks for review and answer it.