func.func @concat_v2(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<6x3xf32> {
  // CHECK: "mhlo.concatenate"({{.*}}) <{dimension = 0 : i64}> : (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<6x3xf32>
  %axis = "tf.Const"() { value = dense<0> : tensor<i64> } : () -> tensor<i64>
  %1 = "tf.ConcatV2"(%arg0, %arg1, %axis) : (tensor<3x3xf32>, tensor<3x3xf32>, tensor<i64>) -> tensor<6x3xf32>
  func.return %1 : tensor<6x3xf32>
}

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>
  }
}

// -----

}

// CHECK-LABEL: fuse_relu_to_add
func.func @fuse_relu_to_add(%arg0: tensor<2x3xf32>, %arg1: tensor<2x3xf32>) -> tensor<2x3xf32> {
  %0 = "tfl.add"(%arg0, %arg1) {fused_activation_function = "NONE"} : (tensor<2x3xf32>, tensor<2x3xf32>) -> tensor<2x3xf32>
  %1 = "tfl.relu_n1_to_1"(%0) : (tensor<2x3xf32>) -> tensor<2x3xf32>
  func.return %1 : tensor<2x3xf32>
  // CHECK: %[[RES:.*]] = tfl.add %arg0, %arg1 {fused_activation_function = "RELU_N1_TO_1"}

    %2 = "quantfork.stats"(%1) {layerStats = dense<[5.00000000e-6, 7.00000000e-1]> : tensor<2xf32>} : (tensor<1x3xf32>) -> tensor<1x3xf32>
    return %2 : tensor<1x3xf32>
  }
// CHECK: func.func private @quantize_dot_general_with_bias_same_shape_fn(%[[ARG_0:.+]]: tensor<1x2xf32>) -> tensor<1x3xf32> attributes {tf._original_func_name = "main_0"}

      %2 = "quantfork.stats"(%1) {layerStats = dense<[5.00000000e-6, 7.00000000e-1]> : tensor<2xf32>} : (tensor<1x3xf32>) -> tensor<1x3xf32>
      return %2 : tensor<1x3xf32>
    }

    func.func private @composite_dot_general_fn(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {

    %5 = "quantfork.qcast"(%4) {volatile} : (tensor<1x3xf32>) -> tensor<1x3x!quant.uniform<i8:f32, 0.13170163023705575:-1>>
    %6 = "quantfork.dcast"(%5) : (tensor<1x3x!quant.uniform<i8:f32, 0.13170163023705575:-1>>) -> tensor<1x3xf32>
    %7 = stablehlo.reshape %6 : (tensor<1x3xf32>) -> tensor<3x1xf32>

    return %1 : tensor<1x3xf32>
  }
  func.func private @composite_matmul_with_bias_fn_1(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>, %arg2: tensor<3xf32>) -> tensor<1x3xf32> attributes {tf_quant.composite_function} {

    version = 5 : i64
  } : (tensor<1x2xf32>, tensor<2x3xf32>) -> tensor<1x3xf32>
  return %0 : tensor<1x3xf32>
}

// CHECK-LABEL: func.func @qdq_for_dot_general_weight_empty
// CHECK-SAME: (%[[ARG_0:.+]]: tensor<1x2xf32>) -> tensor<1x3xf32>
// CHECK: %[[CST:.+]] = "tf.Const"() <{value = dense<3.000000e-01> : tensor<2x3xf32>}> : () -> tensor<2x3xf32>

    %1 = "tf.BiasAdd"(%0, %arg2) <{data_format = "NHWC"}> {device = ""} : (tensor<1x3xf32>, tensor<3xf32>) -> tensor<1x3xf32>
    return %1 : tensor<1x3xf32>
  }

  func.func private @composite_matmul_with_bias_fn_2(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>, %arg2: tensor<3xf32>) -> tensor<1x3xf32> attributes {tf_quant.composite_function} {