// CHECK-DAG:       [[VAL_11:%.*]] = "tf.Const"() <{value = dense<0> : tensor<i32>}> : () -> tensor<i32>
// CHECK:           [[VAL_12:%.*]]:4 = "tf.SplitV"([[VAL_7]], [[VAL_10]], [[VAL_11]]) : (tensor<40x8xf32>, tensor<4xi32>, tensor<i32>) -> (tensor<10x8xf32>, tensor<10x8xf32>, tensor<10x8xf32>, tensor<10x8xf32>)

}

// ================== splitV ========================

LogicalResult UnrollSplitV::matchAndRewrite(TFL::SplitVOp splitv_op,
                                            PatternRewriter& rewriter) const {
  // We need to make sure both splits & split dim are constants.
  auto splits = splitv_op.getSizeSplits().getDefiningOp();
  mlir::DenseIntElementsAttr splits_attr;
  if (!splits || !matchPattern(splits, m_Constant(&splits_attr)))

DECL_CONVERT_OP(BatchMatMulV2);
DECL_CONVERT_OP(BatchMatMulV3);
DECL_CONVERT_OP(MatMul);
DECL_CONVERT_OP(MatrixDiagV2);
DECL_CONVERT_OP(MatrixDiagV3);
DECL_CONVERT_OP(Pack);
DECL_CONVERT_OP(Split);
DECL_CONVERT_OP(SplitV);
DECL_CONVERT_OP(Unpack);
DECL_CONVERT_OP(Conv3D);
DECL_CONVERT_OP(Conv3DBackpropInputV2);

#undef DECL_CONVERT_OP

// Converts any IntegerAttr to an IntegerAttr of an i32 type.

  // CHECK: "tfl.space_to_batch_nd"
}

func.func @split(%arg0: tensor<i32>, %arg1: tensor<1x4x3x3xf32>) -> tensor<1x4x3xf32> {
  %0:3 = "tf.Split"(%arg0, %arg1) : (tensor<i32>, tensor<1x4x3x3xf32>) -> (tensor<1x4x3xf32>, tensor<1x4x3xf32>, tensor<1x4x3xf32>)
  func.return %0#0 : tensor<1x4x3xf32>

  // CHECK-LABEL: split

// tf.Split legalization
//===----------------------------------------------------------------------===//

// -----

// CHECK-LABEL: @split_not_match_dynamic_split_dim_input
func.func @split_not_match_dynamic_split_dim_input(%input: tensor<4x4xf32>, %split_dim: tensor<i32>) -> (tensor<*xf32>, tensor<*xf32>) {
  // CHECK: tf.Split

//===----------------------------------------------------------------------===//
// SplitOp
//===----------------------------------------------------------------------===//

// Verifies the input and split dimension operands for tf.Split/tf.SplitV.
// Writes the split dimension's index (adjusted with input rank) via `dim_index`
// if it's a constant.
template <class Op>
LogicalResult VerifySplitInputAndSplitDim(Op op,

    return success();
  }
};

// Converts the tf.SplitV op into a series of HLO slice ops when the tensor to
// be split has fully static shape and the dimension to split and split sizes
// are constants.
//
// This is similar to the conversion for tf.Split op other than that the size of
// each chunk on the dimension to split is explicitly given as an op operand
// and they are not necessarily the same.
//

    SameOperandsAndResultsScale]> {
  let summary = "Splits a tensor into `num_split` tensors along one dimension.";

  let description = [{
    Splits the `value` tensor along `split_dim` into a number of sub-tensors
    with same shape as the original one, except for `split_dim`. The grouping
    of the resultant sub-tensors is decided by `size-splits`. Same as tf.SplitV.
  }];

  let arguments = (ins

// RUN: tf-opt -tfl-split-merged-operands %s | FileCheck %s

func.func @testSingleLstm(%arg0: tensor<4x4xf32>, %arg1: tensor<4xf32>, %arg2: tensor<4x4x4xf32>) -> tensor<4x4x4xf32> {
  // CHECK-LABEL: testSingleLstm
  // CHECK-DAG:  %[[CST_0:.*]] = "tfl.pseudo_const"() <{value = dense<0.000000e+00> : tensor<4x4xf32>}> : () -> tensor<4x4xf32>
  // CHECK-DAG:  %[[CST_1:.*]] = "tfl.pseudo_const"() <{value = dense<0.000000e+00> : tensor<4x4xf32>}> : () -> tensor<4x4xf32>

        `tf.sparse.cross`/`tf.ragged.cross` directly.
    *   Added additional `standardize` and `split` modes to `TextVectorization`:
        *   `standardize="lower"` will lowercase inputs.
        *   `standardize="string_punctuation"` will remove all punctuation.
        *   `split="character"` will split on every unicode character.
    *   Added an `output_mode` argument to the `Discretization` and `Hashing`