tf.nn.max_pool

max_pool函数经常用到，所以做个记录

参考官方文档

format:max_pool(value, ksize, strides, padding, data_format='NHWC', name=None)

    Args:

      value: A 4-D `Tensor` with shape `[batch, height, width, channels]` and type `tf.float32`.(注意这里数据类型必须是tf.float32，不然会出错)

      ksize: A list of ints that has length >= 4.  The size of the window for each dimension of the input tensor.（一般都是4个元素的list）

      strides: A list of ints that has length >= 4.  The stride of the sliding

        window for each dimension of the input tensor.

      padding: A string, either `'VALID'` or `'SAME'`. The padding algorithm.

        See the @{tf.nn.convolution$comment here}

      data_format: A string. 'NHWC' and 'NCHW' are supported.

      name: Optional name for the operation.