// CHECK:  %5 = "tfl.arg_max"(%arg0, %cst) : (tensor<4x32x256xf32>, tensor<1xi32>) -> tensor<4x32xi32>
  // CHECK:  return %4, %5 : tensor<4x32xf32>, tensor<4x32xi32>
}

// CHECK-LABEL: func @convert_argmax_constant
func.func @convert_argmax_constant(%arg0: tensor<2x2x4xf32>) -> (tensor<2x2xf32>, tensor<2x2xi32>) {
  %0 = mhlo.constant dense<0xFF800000> : tensor<f32>

func.func @concat_v2(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<6x3xf32> {
  %2 = "mhlo.concatenate"(%arg0, %arg1) <{dimension = 0 : i64}> : (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<6x3xf32>
  func.return %2 : tensor<6x3xf32>
}

// CHECK-LABEL:   func @concat_v2_1d_axis(
// CHECK-SAME:                            %[[VAL_0:.*]]: tensor<3x3xf32>,
// CHECK-SAME:                            %[[VAL_1:.*]]: tensor<3x3xf32>) -> tensor<3x6xf32> {

// CHECK: return

  func.func private @composite_dot_general_fn(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
      %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
      return %0 : tensor<1x3xf32>
  }
}

// -----

    %3 = stablehlo.concatenate %2, %1, dim = 0 : (tensor<1x3xf32>, tensor<1x3xf32>) -> tensor<2x3xf32>
    return %3 : tensor<2x3xf32>
  }
  func.func private @composite_dot_general_fn_1(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module, tf_quant.composite_function} {

    return %1 : tensor<1x3xf32>
  }

  func.func private @composite_dot_general_fn(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
    %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x2xf32>, tensor<2x3xf32>) -> tensor<1x3xf32>
    return %0 : tensor<1x3xf32>
  }
}

  func.return %0: tensor<1x2x4xf32>
}

func.func @test_batch_matmulV2_adj_to_einsum(%arg0: tensor<1x3x2xf32>, %arg1: tensor<3x4xf32>) -> tensor<1x2x4xf32> {
  // CHECK: %[[RES_EINSUM:[0-9]*]] = "tf.Einsum"(%arg0, %arg1) <{equation = "...km,...kn->...mn"}> : (tensor<1x3x2xf32>, tensor<3x4xf32>) -> tensor<1x2x4xf32>
  // CHECK: return %[[RES_EINSUM]] : tensor<1x2x4xf32>

      %2 = "tf.XlaCallModule"(%arg0, %1, %0) <{Sout = [#tf_type.shape<?x2>], module = "", version = 9 : i64}> {_original_entry_function = "composite_fn_1"} : (tensor<?x2xf32>, tensor<2x2xf32>, tensor<2xf32>) -> tensor<?x2xf32>
      return %2 : tensor<?x2xf32>
    }

    // CHECK-NEXT:  %[[RES:.*]] = "tf.SelectV2"(%[[PRED]], %[[SCALED_GRADIENTS]], %[[SELU_GRAD_VALUE]]) : (tensor<4x8xi1>, tensor<4x8xf32>, tensor<4x8xf32>) -> tensor<4x8xf32>
    // CHECK-NEXT:  return %[[RES]] : tensor<4x8xf32>
    %2 = "tf.SeluGrad"(%gradients, %features) : (tensor<4x8xf32>, tensor<4x8xf32>) -> tensor<4x8xf32>
    func.return %2 : tensor<4x8xf32>
}

  // If operand has the same shape as a result, we can fold it.
  %3 = "tf.AddV2"(%arg1, %0) : (tensor<4x2xf32>, tensor<1x2xf32>) -> tensor<4x2xf32>
  %4 = "tf.AddV2"(%0, %arg1) : (tensor<1x2xf32>, tensor<4x2xf32>) -> tensor<4x2xf32>

  // CHECK: %[[CONST:.*]] = "tf.Const"()
  // CHECK-DAG: %[[ADD1:.*]] = "tf.AddV2"(%arg0, %[[CONST]])

    %12 = "tfl.logistic"(%11) : (tensor<4x2xf32>) -> tensor<4x2xf32>
    %13 = tfl.mul %arg4, %12 {fused_activation_function = "NONE"} : tensor<4x2xf32>
    %14 = "tfl.relu"(%10#1) : (tensor<4x2xf32>) -> tensor<4x2xf32>
    %15 = "tfl.logistic"(%10#0) : (tensor<4x2xf32>) -> tensor<4x2xf32>
    %16 = tfl.mul %15, %14 {fused_activation_function = "NONE"} : tensor<4x2xf32>