// not contributing to the dimentionality. This will run if the rank>6
// Pattern will convert-
// %0 = "tfl.transpose"(%arg0, %cst) : (tensor<56x8x56x1x1x1x7xf32>, tensor<7xi32>) -> tensor<1x1x8x56x56x7x1xf32>
// to-
// %0 = "tfl.reshape"(%arg0, %cst) : (tensor<56x8x56x1x1x1x7xf32>, tensor<4xi32>) -> tensor<56x8x56x7xf32>
// %1 = "tfl.transpose"(%0, %cst_0) : (tensor<56x8x56x7xf32>, tensor<4xi32>) -> tensor<8x56x56x7xf32>

func.func @FuseTransposeFCRhsToBatchMatmul(%arg0: tensor<16x1024xf32>, %arg1: tensor<1024x128xf32>, %arg2: none) -> tensor<16x128xf32> {
  %cst = arith.constant dense<[1, 0]> : tensor<2xi32>
  %0 = "tfl.transpose"(%arg1, %cst) : (tensor<1024x128xf32>, tensor<2xi32>) -> tensor<128x1024xf32>
  // CHECK: "tfl.batch_matmul"(%arg0, %arg1)

// with `size` number of values.
bool IsIotaAttr(ArrayRef<int64_t> arr, int64_t size);

// Returns a DenseIntElementsAttr for a permutation and the shape after
// applying the permutation to a given shape through a transpose.
PermutationAndShape GetPermutationAndTransposedShape(
    llvm::ArrayRef<int64_t> permutation_array, ShapedType input_type,
    ConversionPatternRewriter& rewriter);

// Create a single const integer.

            .transpose()
            .each { actualItem, expectedItem ->
                checker(actualItem, expectedItem)
            }
    }

    static void checkMap(Map<?, ?> actual, Map<?, ?> expected, Closure checker) {
        assert actual.size() == expected.size()
        [actual, expected].collect { new ArrayList<>(it.entrySet()) }
            .transpose()

  pm.addNestedPass<func::FuncOp>(createNchwConvolutionToNhwcPass());

  // Recursively push down the `stablehlo.transpose` ops for activations
  // generated by the `NchwConvolutionToNhwc` pass.
  pm.addNestedPass<func::FuncOp>(createDeferActivationTransposePass());

  // Folds `stablehlo.constant`->`stablehlo.transpose` patterns, which is often
  // generated as by-products after optimizing dimension numbers (e.g.

 Relu 49
 ReluGrad 49
 Reshape 2
 ResourceApplyKerasMomentum 161
 ShapeN 50
 Softmax 1
 SparseSoftmaxCrossEntropyWithLogits 1
 Square 55
 Squeeze 1
 Sub 106
 Sum 57
 Tile 1

func.func @main(%arg0: tensor<5x7xf32>) -> tensor<5x7xf32> {
  func.return %arg0: tensor<5x7xf32>
// CHECK-LABEL: main
// CHECK: return %arg0 : tensor<5x7xf32>
}

// - transpose
//
func.func @transpose_2d(%arg0: tensor<2x3xf32>) -> tensor<3x2xf32> {
  %0 = "mhlo.transpose"(%arg0) <{permutation = dense<[1, 0]> : tensor<2xi64>}> : (tensor<2x3xf32>) -> tensor<3x2xf32>
  func.return %0 : tensor<3x2xf32>

// CHECK-LABEL:   transpose_2d

    reserve_spaces[1] = op.input(4);  // pop_var
    if (data_format == "NCHW") {
      x = Transpose(scope, x, {0, 2, 3, 1});
      grad_y = Transpose(scope, grad_y, {0, 2, 3, 1});
    } else if (data_format == "NCDHW") {
      x = Transpose(scope, x, {0, 2, 3, 4, 1});
      grad_y = Transpose(scope, grad_y, {0, 2, 3, 4, 1});
    }

    StringPiece target_data_format;

 Mul 218
 Pad 2
 ReadVariableOp 538
 Relu 49
 ReluGrad 49
 Reshape 2
 ResourceApplyKerasMomentum 161
 Slice 1
 Softmax 1
 SparseSoftmaxCrossEntropyWithLogits 1
 Squeeze 1
 Sum 1
 Tile 1
 Transpose 1
cluster 1 size 815
 AddN 1
 AssignAddVariableOp 1
 AssignSubVariableOp 106
 Const 220
 DivNoNan 1
 Identity 1
 Mul 161
 ReadVariableOp 106
 Square 55
 Sub 106

  %2 = "tf.Transpose"(%0, %cst_0) {device = ""} : (tensor<1x3x4x3xbf16>, tensor<4xi32>) -> tensor<1x3x3x4xbf16>
  %3 = "tf.TPUReplicatedInput"(%2) {device = "", index = -1 : i64, is_mirrored_variable = false, is_packed = false} : (tensor<1x3x3x4xbf16>) -> tensor<1x3x3x4xbf16>