batch_to_space changes for TF 2.0 API
This change renames batch_to_space_nd to batch_to_space in TF 2.0. Arguments are re-ordered and the the block_size int argument has been to block_shape of type Tensor. Users may still use an int block_shape, for compatibility purposes those are silently converted to a Tensor. PiperOrigin-RevId: 222411473
Loading
Please sign in to comment