// CHECK: %[[v1:.*]] = "tf.BatchMatMulV2"(%arg0, %[[v0]]) <{adj_x = false, adj_y = false}> : (tensor<2x5x7x3xf32>, tensor<2x5x3x13xf32>) -> tensor<2x5x7x13xf32>
  // CHECK: %[[v2:.*]] = "tf.Transpose"(%[[v1]], %[[cst]]) : (tensor<2x5x7x13xf32>, tensor<4xi32>) -> tensor<2x7x5x13xf32>
  // CHECK: return %[[v2]] : tensor<2x7x5x13xf32>
}

  %cst = "tf.Const"() {value = dense<2.000000e+00> : tensor<2x3x3x1xf32>} : () -> tensor<2x3x3x1xf32>
  %cst_0 = "tf.Const"() {value = dense<0.400000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
  %cst_1 = "tf.Const"() {value = dense<0.500000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
  %identity = "tf.Identity"(%cst) : (tensor<2x3x3x1xf32>) -> tensor<2x3x3x1xf32>

//==== V1 tests ====

func.func @batchMatMulTwoDim(%arg0: tensor<2x3x4x5xf32>, %arg1: tensor<2x3x5x6xf32>) -> tensor<2x3x4x6xf32> {
  %0 = "tf.BatchMatMul"(%arg0, %arg1) : (tensor<2x3x4x5xf32>, tensor<2x3x5x6xf32>) -> tensor<2x3x4x6xf32>
  func.return %0 : tensor<2x3x4x6xf32>

  // CHECK-LABEL: batchMatMulTwoDim

        %cst = "tf.Const"() {value = dense<2.000000e+00> : tensor<2x3x3x1xf32>} : () -> tensor<2x3x3x1xf32>
        %cst_0 = "tf.Const"() {value = dense<0.400000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
        %cst_1 = "tf.Const"() {value = dense<0.500000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
        %w = "tf.Mul"(%cst, %arg1) : (tensor<2x3x3x1xf32>, tensor<f32>) -> tensor<2x3x3x1xf32>

    %12 = "quantfork.qcast"(%11) {volatile} : (tensor<2x3x1x3xf32>) -> tensor<2x3x1x3x!quant.uniform<i8:f32, 3.000000e-01:1>>
    %13 = "quantfork.dcast"(%12) : (tensor<2x3x1x3x!quant.uniform<i8:f32, 3.000000e-01:1>>) -> tensor<2x3x1x3xf32>
    return %13 : tensor<2x3x1x3xf32>
  }

  // CHECK: quantized_dot_general_fn_1

      stablehlo.return %2 : tensor<f32>
  }) {padding = dense<[[0, 0], [1, 1], [1, 1], [0, 0]]> : tensor<4x2xi64>, window_dimensions = array<i64: 1, 3, 3, 1>} : (tensor<2x3x1x3xf32>, tensor<f32>) -> tensor<2x3x1x3xf32>
  return %1 : tensor<2x3x1x3xf32>
}

// -----

// CHECK-LABEL: @bitcast_convert_i32_f32(%arg0: tensor<1x256128xi32>) -> tensor<1x256128xbf16>

    %0 = "tf.DepthwiseConv2dNative"(%arg0, %cst_0) {data_format = "NHWC", dilations = [1, 1, 1, 1], explicit_paddings = [], padding = "SAME", strides = [1, 2, 2, 1]} : (tensor<1x3x4x3xf32>, tensor<2x3x3x1xf32>) -> tensor<1x2x2x3xf32>

// CHECK: return %1 : tensor<5x2x3x4xf32>

// -----

// CHECK-LABEL: pushTposeBcastNoChange
func.func @pushTposeBcastNoChange(%arg0: tensor<2x3x4x1xf32>) -> tensor<5x2x3x4xf32> {
  %perm = arith.constant dense<[3, 0, 1, 2]> : tensor<4xi32>
  %0 = "tfl.transpose"(%arg0, %perm) : (tensor<2x3x4x1xf32>, tensor<4xi32>) -> tensor<1x2x3x4xf32>
  %cst = arith.constant dense<1.0> : tensor<5x2x3x4xf32>

    %cst_1 = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x3x1xf32>} : () -> tensor<2x3x3x1xf32>
    %cst_2 = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>
    %0 = "tf.PartitionedCall"(%arg0, %cst_1) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_depthwise_conv2d_fn} : (tensor<1x3x4x3xf32>, tensor<2x3x3x1xf32>) -> tensor<*xf32>

  %cst_1 = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x3x1xf32>} : () -> tensor<2x3x3x1xf32>
  %0 = "tf.DepthwiseConv2dNative"(%arg0, %cst_1) {
    data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [],
    padding = "SAME", strides = [1, 1, 2, 1]
  } : (tensor<1x3x4x3xf32>, tensor<2x3x3x1xf32>) -> tensor<*xf32>