// CHECK-DAG: %[[input_18:.*]] = "quantfork.stats"({{.*}}) <{layerStats = dense<[-8.000000e-01, 1.600000e+00]> : tensor<2xf32>}> : (tensor<1x4xf32>) -> tensor<1x4xf32>
// CHECK-DAG: %[[input_19:.*]] = "quantfork.stats"({{.*}}) <{layerStats = dense<[-2.000000e+00, 4.000000e+00]> : tensor<2xf32>}> : (tensor<1x2xf32>) -> tensor<1x2xf32>

// CHECK: "tfl.unidirectional_sequence_lstm"({{.*}}, %[[input_18]], %[[input_19]], %{{[0-9]+}}, %{{[0-9]+}}, %{{[0-9]+}}, %{{[0-9]+}})

    %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32>
    %cst_0 = "tf.Const"() {value = dense<[-1, 10]> : tensor<2xi32>} : () -> tensor<2xi32>
    %1 = "tf.MatMul"(%arg0, %arg1) {
      transpose_a = false, transpose_b = false

    module attributes {} {
      func.func @main(%arg0: tensor<1x4xf32>) -> tensor<1x3xf32> attributes {} {
        %0 = stablehlo.constant dense<1.0> : tensor<4x3xf32>
        %1 = stablehlo.dot_general %arg0, %0, contracting_dims = [1] x [0], precision = [DEFAULT, DEFAULT] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
        return %1 : tensor<1x3xf32>
      }
    }
  )mlir");
  ASSERT_TRUE(module_op);

  // CHECK-SAME: -> tensor<!tf_type.variant<tensor<16x1xf32>>>
  func.func @while_variant(%arg0: tensor<!tf_type.variant<tensor<16x1xf32>>>) -> tensor<!tf_type.variant> {
    // CHECK: tf.While
    // CHECK-SAME: -> tensor<!tf_type.variant<tensor<16x1xf32>>>
    %0 = "tf.While"(%arg0) {cond = @variant_cond_func, body = @variant_body_func, is_stateless = true} : (tensor<!tf_type.variant<tensor<16x1xf32>>>) -> tensor<!tf_type.variant>

  %elem_2 = "tf._SecondOp"() : () -> tensor<16x1xf32>
  %tl_set_item2 = "tf.TensorListSetItem"(%tl, %one, %elem_2) : (tensor<!tf_type.variant<tensor<?x1xf32>>>, tensor<i32>, tensor<16x1xf32>) -> tensor<!tf_type.variant<tensor<?x1xf32>>>
  func.return
}

// CHECK: use_two_sep_ops_empty_tensor_list

  %1 = "tf.ReadVariableOp"(%0) {device = "/CPU:0"} : (tensor<!tf_type.resource<tensor<1x1xf32>>>) -> tensor<1x1xf32>
  %2:2 = "tf.Split"(%cst, %arg0) {device = "/CPU:0"} : (tensor<i32>,  tensor<1x4xf32>) -> (tensor<1x2xf32>, tensor<1x2xf32>)
  %3:2 = "tf.Split"(%cst, %2#0) {device = "/CPU:0"} : (tensor<i32>,  tensor<1x2xf32>) -> (tensor<1x1xf32>, tensor<1x1xf32>)

// -----

// CHECK-LABEL: float_matmul
func.func @float_matmul(
  %arg0: tensor<1x10xf32>, %arg1: tensor<10x10xf32>) -> (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) {
  %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32>
  %0 = "tf.MatMul"(%arg0, %arg1) {
    transpose_a = false, transpose_b = false
  } : (tensor<1x10xf32>, tensor<10x10xf32>) -> tensor<*xf32>

  %5 = stablehlo.broadcast_in_dim %1, dims = [2] : (tensor<64xf32>) -> tensor<1x1x64xf32>
  %6 = stablehlo.add %4, %5 : tensor<1x1x64xf32>
  %7 = stablehlo.clamp %2, %6, %3 : tensor<1x1x64xf32>
  func.return %7: tensor<1x1x64xf32>
}
// CHECK: %[[CONST_0:.*]] = stablehlo.constant dense<2.000000e+00>
// CHECK: %[[CONST_1:.*]] = stablehlo.constant dense<2.000000e+00>

}

// CHECK-LABEL: @broadcast_batch_matmul_v2_rhs
func.func @broadcast_batch_matmul_v2_rhs(%arg0: tensor<17x17x17xf32>, %arg1: tensor<17x24xf32>) -> tensor<17x17x24xf32> {
  %cst = arith.constant dense<[17, 17, 24]> : tensor<3xi64>
  %0 = "tf.BroadcastTo"(%arg1, %cst) : (tensor<17x24xf32>, tensor<3xi64>) -> tensor<17x17x24xf32>

  %0:2 = "tf.Split"(%cst, %input) : (tensor<i32>, tensor<4x?x4xf32>) -> (tensor<2x?x4xf32>, tensor<2x?x4xf32>)
  func.return %0#0, %0#1 : tensor<2x?x4xf32>, tensor<2x?x4xf32>
}

// -----

// CHECK-LABEL: @split_match_and_split_into_two
func.func @split_match_and_split_into_two(%input: tensor<4x6xf32>) -> (tensor<2x6xf32>, tensor<2x6xf32>) {