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>

  func.return %17 : tensor<1x2x3xf32>

  // CHECK: %[[NONE:.*]] = "tfl.no_value"() <{value}> : () -> none
  // CHECK: %[[DQ_1:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>
  // CHECK: %[[DQ_2:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>
  // CHECK: %[[DQ_3:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>

# CHECK:           %[[VAL_9:.*]] = "tfl.transpose"(%[[VAL_1]], %[[VAL_2]]) : (tensor<3x7xf32>, tensor<2xi32>) -> tensor<7x3xf32>
# CHECK:           %[[VAL_10:.*]] = "tfl.fully_connected"(%[[VAL_7]]#0, %[[VAL_9]], %[[VAL_3]]) <{fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"}> : (tensor<1x5x3xf32>, tensor<7x3xf32>, none) -> tensor<5x7xf32>

  }
  tfr.return %res : !tfr.tensor
}

// CHECK-LABEL: pack_one
func.func @pack_one(%arg0: tensor<2x3xf32>) -> tensor<1x2x3xf32> {
  %0 = "tf.MyPack"(%arg0) {N=1:i32, axis=0:i32} : (tensor<2x3xf32>) -> tensor<1x2x3xf32>
  func.return %0 : tensor<1x2x3xf32>

// CHECK-NEXT: %[[AXIS:.*]] = arith.constant 0 : i32
// CHECK-NEXT: %[[CAST:.*]] = "tfr.cast"(%arg0) : (tensor<2x3xf32>) -> !tfr.tensor

  // CHECK: device = "gpu"
  %3 = "tf.Relu"(%2) {T = f32, _output_shapes = ["tfshape$dim { size: 3 } dim { size: 3 }"]} : (tensor<3x3xf32>) -> tensor<3x3xf32>
  func.return %3 : tensor<3x3xf32>

  // CHECK-NEXT: [[out_ch:%.*]] = tfrt.merge.chains [[ch]], [[ch1]] : !tfrt.chain, !tfrt.chain
  // CHECK-NEXT: tfrt.return [[out_ch]], [[r2]] : !tfrt.chain, !tfrt_fallback.tf_tensor
  func.return %6#0 : tensor<3x3xf32>
}

  %cst = arith.constant dense<1.0> : tensor<4xf32>
  %cst_3 = arith.constant dense<2.0> : tensor<4x3x3x3xf32>
  %0 = "tfl.conv_2d"(%arg0, %cst_3, %cst) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU6", padding = "VALID", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<?x19x19x3xf32>, tensor<4x3x3x3xf32>, tensor<4xf32>) -> tensor<?x9x9x4xf32>
  func.return %0 : tensor<?x9x9x4xf32>

  func.return %1 : tensor<3x3xf32>
}

// CHECK-LABEL: func @gpu_device
func.func @gpu_device(%arg0: tensor<3x1xf32>, %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>>) -> tensor<3x3xf32> {
  // CHECK: {{%.*}} = corert.get_op_handler %arg0 "/device:GPU:0"

      -> (tensor<3x3xf32> {tf_saved_model.index_path = []}) {
  // CHECK: {{%.*}} = corert.get_op_handler %arg0 "/device:GPU:0"
  %2 = "tf.MatMul"(%arg0, %arg1) {T = f32, _output_shapes = ["tfshape$dim { size: 3 } dim { size: 3 }"], device = "/device:GPU:0", transpose_a = false, transpose_b = false} : (tensor<3x1xf32>, tensor<1x3xf32>) -> tensor<3x3xf32>
  func.return %2 : tensor<3x3xf32>

// CHECK-LABEL: @add_f32(%arg0: tensor<3x3xbf16>, %arg1: tensor<3x3xbf16>) -> tensor<3x3xbf16>
func.func @add_f32(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<3x3xf32> {
  // CHECK-NOT: f32
  // CHECK: stablehlo.add
  %0 = stablehlo.add %arg0, %arg1: (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<3x3xf32>
  return %0 : tensor<3x3xf32>
}

// -----

// CHECK-LABEL: @add_f64(%arg0: tensor<3x3xbf16>, %arg1: tensor<3x3xbf16>) -> tensor<3x3xbf16>