// CHECK-DAG: [[RESULT:%.+]] = "tf.Reshape"([[PERMUTED]], [[OUTPUT_SHAPE]])
  // CHECK-DAG: return [[RESULT]]
  %0 = "tf.SpaceToBatchND"(%input, %block_shape, %paddings) : (tensor<3x5x7x10xf32>, tensor<2xi64>, tensor<2x2xi64>) -> tensor<?x?x?x10xf32>
  func.return %0 : tensor<?x?x?x10xf32>
}

// Verify SpaceToBatchND with input tensor of element type f16. This test case is derived from 'fourdim_space_to_batch_nd'. It checks the output

                                            axis_value);
    return success();
  }
};

// Lowers SpaceToBatchND by reducing to reshape(transpose(reshape(pad(input)))).
//
// Before rewrite:
//   output = SpaceToBatchND(input, block_shape, paddings)
// Let:
//   [batch] + spatial_shape + remaining_shape = input.shape
//   M = spatial_shape.rank
// After rewrite:

  }

  if (failed(ConvertTf2XlaOps(func, ctx))) {
    signalPassFailure();
    return;
  }

  // This pattern will try to identify and optimize for dilated convolution.
  // e.g. Patterns like "SpaceToBatchND -> Conv2D -> BatchToSpaceND" will be
  // replaced with a single Conv op with dilation parameter.
  patterns.add<ConvertTFDilatedConvOp<TF::Conv2DOp>, FusedBatchNormV3Pat,

      "SegmentSumV2",
      "SegmentProdV2",
      "SegmentMinV2",
      "SegmentMaxV2",
      "SelfAdjointEigV2",
      "SoftmaxCrossEntropyWithLogits",
      "SpaceToBatch",
      "SpaceToBatchND",
      "SpaceToDepth",
      "SparseMatMul",
      "SparseToDense",
      "StackCloseV2",
      "StackPopV2",
      "StackPushV2",
      "StackV2",
      "StatefulPartitionedCall",