# CHECK-LABEL:      func @key(
# CHECK-SAME:   %[[ARG0:.*]]: tensor<3x5xf32> {tf_saved_model.index_path = ["y"]}
# CHECK-SAME:   %[[ARG1:.*]]: tensor<5x3xf32> {tf_saved_model.index_path = ["x"]}
# CHECK-SAME:                  tensor<3x3xf32> {tf_saved_model.index_path = ["t"]}
# CHECK-SAME:                  tensor<5x5xf32> {tf_saved_model.index_path = ["s"]}
# CHECK-SAME: attributes {{.*}} tf_saved_model.exported_names = ["key"]

    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);

func.func @testSoftmaxCrossEntropyWithLogits(%arg0: tensor<2x3xf32>, %arg1: tensor<3xf32>) -> (tensor<3xf32>, tensor<2x3xf32>) {
  %0:2 = "tf.SoftmaxCrossEntropyWithLogits"(%arg0, %arg1) : (tensor<2x3xf32>, tensor<3xf32>) -> (tensor<3xf32>, tensor<2x3xf32>)
  func.return %0#0, %0#1 : tensor<3xf32>, tensor<2x3xf32>
}

// -----

func.func @testSoftmaxCrossEntropyWithLogits(%arg0: tensor<2xf32>, %arg1: tensor<3xf32>) -> (tensor<3xf32>, tensor<3xf32>) {

      %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>
    }

  %8 = "tfl.concatenation"(%2, %0) {axis = -1 : i32, fused_activation_function = "NONE"} : (tensor<1x1xf32>, tensor<1x1xf32>) -> tensor<1x2xf32>
  %9 = "quantfork.stats"(%8) {layerStats = dense<[-0.488159984, 0.189515018]> : tensor<2xf32>} : (tensor<1x2xf32>) -> tensor<1x2xf32>
  %10 = "tfl.concatenation"(%9, %7) {axis = -1 : i32, fused_activation_function = "NONE"} : (tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<1x4xf32>

// CHECK:           }) : (tensor<1x?xf32>, tensor<1x0xf32>, tensor<1x0xf32>, tensor<1x0xf32>, tensor<1x0xf32>, tensor<1x3xf32>, tensor<1x3xf32>, tensor<1x3xf32>, tensor<1x3xf32>, none, none, none, tensor<1xf32>, tensor<1xf32>, tensor<1xf32>, tensor<1xf32>, tensor<3x1xf32>, tensor<3xf32>, tensor<1x3xf32>, tensor<1x1xf32>, none, none, none, none) -> tensor<1x3xf32>
// CHECK:           [[VAL_50:%.*]] = tensor.cast [[VAL_51:%.*]] : tensor<1x3xf32> to tensor<1x?xf32>

func.func @testSubOfZero(%arg0: tensor<?x1xf32>, %arg1: tensor<4x1xf32>) -> (tensor<?x1xf32>, tensor<4x1xf32>) {
  %0 = "tf.Const"() {value = dense<0.0> : tensor<f32>} : () -> tensor<f32>
  %1 = "tf.Sub"(%arg0, %0) : (tensor<?x1xf32>, tensor<f32>) -> tensor<?x1xf32>
  %2 = "tf.Sub"(%arg1, %0) : (tensor<4x1xf32>, tensor<f32>) -> tensor<4x1xf32>
  func.return %1, %2: tensor<?x1xf32>, tensor<4x1xf32>

// CHECK: return %arg0, %arg1
}

  // CHECK-DAG: %[[EXP:.*]] = "tf.Exp"(%[[SHIFTED]]) : (tensor<2x3xf32>) -> tensor<2x3xf32>
  // CHECK-DAG: %[[SUM:.*]] = "tf.Sum"(%[[EXP]], %[[AXIS]]) <{keep_dims = true}> : (tensor<2x3xf32>, tensor<1xi64>) -> tensor<2x1xf32>
  // CHECK-DAG: %[[RESULT:.*]] = "tf.Div"(%[[EXP]], %[[SUM]]) : (tensor<2x3xf32>, tensor<2x1xf32>) -> tensor<2x3xf32>
  // CHECK: return %[[RESULT]]

}

func.func @addN(%arg0: tensor<2x3xi32>, %arg1: tensor<2x3xi32>, %arg2: tensor<2x3xi32>) -> tensor<2x3xi32> {
  %0 = "tf.AddN"(%arg0, %arg1, %arg2) : (tensor<2x3xi32>, tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
  func.return %0 : tensor<2x3xi32>

// CHECK-LABEL: addN
// CHECK:  "tfl.add_n"(%arg0, %arg1, %arg2) : (tensor<2x3xi32>, tensor<2x3xi32>, tensor<2x3xi32>) -> tensor<2x3xi32>
// CHECK:  return
}

  %outputs =  "tf.Const"() {device = "/job:localhost/replica:0/task:0/device:CPU:0", value = dense<0> : tensor<i32>} : () -> tensor<i32>...