When I use layers related to DepthwiseConv2d operation, such as tf.keras.layers.SeparableConv2d, exception occurs like that: 'Current implementation only supports equal length strides in the row and column dimensions. [Op: DepthwiseConv2dNative]'. It means that i cannot use parameter 'strides' like [1, 2], however, that conflicts with the documentation, which allows 'strides' with list format without mentioning that elements within the list should be the same.
I'm using TensorFlow 2.0.0, and I guess this bug exisits in 3D scenariothe and further edition.