.. meta:: :description: Backward pass of separable 3D average pooling. :date-modified: 10/19/2026 .. currentmodule:: nkilib.experimental.pooling Average Pooling 3D Backward Kernel API Reference ================================================ Backward pass of separable 3D average pooling. Spreads each output gradient back over the input window it was averaged from, applying the 1/prod(kernel_size) scale, using a separable per-axis spread across D, H and W. Background ----------- The ``avg_pooling_3d_backward`` kernel computes the backward pass of 3D average pooling. It spreads each output gradient back over its input window with the ``1/prod(kernel_size)`` scale, one axis at a time across D, H and W. API Reference -------------- **Source code for this kernel API can be found at**: `avg_pooling_3d_backward.py `_ avg_pooling_3d_backward ^^^^^^^^^^^^^^^^^^^^^^^ .. py:function:: avg_pooling_3d_backward(grad_output: nl.NkiTensor, input_tensor: nl.NkiTensor, kernel_size: Union[int, tuple], stride: Optional[Union[int, tuple]] = None, padding: Union[int, tuple] = 0, data_format: str = 'NCDHW', output_format: Optional[str] = None) -> nl.NkiTensor Backward pass of separable 3D average pooling. :param grad_output: Gradient w.r.t. the forward output, in HBM, output_format layout, forward-output shape. :type grad_output: ``nl.NkiTensor`` :param input_tensor: The forward input, data_format layout. Only its shape and dtype are read an average pool's gradient does not depend on the forward values. :type input_tensor: ``nl.NkiTensor`` :param kernel_size: Pooling window (kD, kH, kW). :type kernel_size: ``Union[int, tuple]`` :param stride: Stride (sD, sH, sW); defaults to kernel_size. :type stride: ``Optional[Union[int, tuple]]`` :param padding: Implicit zero padding (pD, pH, pW). :type padding: ``Union[int, tuple]`` :param data_format: Layout of input_tensor and grad_input, "NCDHW" or "NDHWC". :type data_format: ``str`` :param output_format: Layout of grad_output; defaults to data_format. :type output_format: ``Optional[str]`` :return: Gradient w.r.t. the forward input, in HBM, data_format layout, same spatial extents as input_tensor. :rtype: ``nl.NkiTensor`` **Dimensions**: * B: Batch size * C: Number of channels * D: Input depth (spatial) * H: Input height (spatial) * W: Input width (spatial) * Do: Output depth (pooled) * Ho: Output height (pooled)