MinPool

Pools the input by taking the minimum of a window slice.

Abstract Signature:

MinPool(inputs: Tensor, window_shape: Union[int, Tuple[int, ...]], strides: Union[int, Tuple[int, ...]], padding: str = VALID)

PyTorchFlax NNX

PyTorch

API: —

Strategy: Macro '-torch.nn.functional.max_pool2d(-{inputs}, kernel_size={window_shape}, stride={strides})'

Flax NNX

API: flax.nnx.min_pool

Strategy: Direct Mapping

Official Docs ↗