<< kernel_output_features << ") to be divisible by "
                            << "feature_group_count (value "
                            << feature_group_count_val << ").\n";
  }
  if (input_batch % batch_group_count != 0) {
    return op.emitOpError()
           << "Expected input batch dimension (value " << input_batch
           << " ) to be divisible by batch_group_count (value "

    // all spatial dimensions.
    const int64_t filter_channels = GetDimSize(filter_ty, num_spatial_dims);
    // TensorFlow convolution op verifies that the number of input channels is
    // divisible by the number of filter channels.
    // For depthwise convolution the feature_group_count argument would be set
    // to the input feature dimension.
    const int64_t feature_group_count =

  }];

  let arguments = (ins
    Arg<TF_Tensor, [{4-D tensor with shape
`[batch*block_size*block_size, height_pad/block_size, width_pad/block_size,
  depth]`. Note that the batch size of the input tensor must be divisible by
`block_size * block_size`.}]>:$input,
    Arg<TF_I32OrI64Tensor, [{2-D tensor of non-negative integers with shape `[2, 2]`. It specifies
how many elements to crop from the intermediate result across the spatial