%3 = "tf.ReadVariableOp"(%arg2) : (tensor<!tf_type.resource<tensor<5xf32>>>) -> tensor<5xf32>
  %4 = "tf.ReadVariableOp"(%arg1) : (tensor<!tf_type.resource<tensor<3x3x1x5xf32>>>) -> tensor<3x3x1x5xf32>
  %5 = "tf.ReadVariableOp"(%arg3) : (tensor<!tf_type.resource<tensor<3920x10xf32>>>) -> tensor<3920x10xf32>

// CHECK-LABEL: strips_attributes
// CHECK-NOT: tf
func.func @strips_attributes(%arg0: tensor<32x28x28x1xf32> {tf._user_specified_name = "x"},
                             %arg1: tensor<3x3x1x5xf32> {tf._user_specified_name = "w1"},
                             %arg2: tensor<5xf32> {tf._user_specified_name = "b1"},
                             %arg3: tensor<3920x10xf32> {tf._user_specified_name = "w2"},

  %w = arith.constant dense<127.0> : tensor<3x3x3x3xf32>
  %b = arith.constant dense<0.0> : tensor<3xf32>
  %conv = "tfl.conv_2d"(%arg0, %w, %b) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x5x5x3xf32>, tensor<3x3x3x3xf32>, tensor<3xf32>) -> tensor<1x5x5x3xf32>
  func.return %conv : tensor<1x5x5x3xf32>

// RUN: tf-opt -split-input-file -verify-diagnostics -tf-unroll-batch-matmul %s | FileCheck %s

//==== 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_0 = arith.constant dense<[1.0, 2.0, 3.0]> : tensor<3xf32>
  %w = arith.constant dense<2.0> : tensor<3x3x3x3xf32>
  %q = "tfl.quantize"(%w) {qtype = tensor<3x3x3x3x!quant.uniform<i8:f32, 0.1:1>>} : (tensor<3x3x3x3xf32>) -> tensor<3x3x3x3x!quant.uniform<i8:f32, 0.1:1>>
  %dq = "tfl.dequantize"(%q) : (tensor<3x3x3x3x!quant.uniform<i8:f32, 0.1:1>>) -> tensor<3x3x3x3xf32>

// CHECK: %0 = tfl.sub %arg0, %cst_0 {fused_activation_function = "NONE"} : tensor<2x3x4x5xf32>
// CHECK: %1 = "tfl.transpose"(%0, %cst) : (tensor<2x3x4x5xf32>, tensor<4xi32>) -> tensor<5x2x3x4xf32>
// CHECK: return %1 : tensor<5x2x3x4xf32>

// -----

// CHECK-LABEL: permNotConstNoChange
func.func @permNotConstNoChange(%arg0: tensor<2x3x4x5xf32>, %perm: tensor<4xi32>) -> tensor<5x2x3x4xf32> {

    %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

    return %5 : tensor<1x3x1x1xf32>
  }
  func.func private @composite_gather_fn_1(%arg0: tensor<1x3x1x1xf32>, %arg1: tensor<1xi32>, %arg2: tensor<i32>) -> tensor<1x3x1x1xf32> attributes {tf_quant.composite_function} {

    : (tensor<*x!tf_type.resource>) -> (tensor<3x3x1x32xf32>, tensor<3x3x1x32xf32>)
  "tf_device.launch"() ({
    "tf.TPUCompileSucceededAssert"(%compile#0) : (tensor<!tf_type.string>) -> ()
    tf_device.return
  }) {device = "/device:CPU:0"} : () -> ()
  %execute0 = "tf_device.launch"() ({
    %3 = "tf.TPUExecute"(%2#0, %2#1, %compile#1)
      : (tensor<3x3x1x32xf32>, tensor<3x3x1x32xf32>, tensor<2x!tf_type.string>) -> tensor<i32>

  %9 = "tfl.quantize"(%8) {qtype = tensor<1x3x1x1x!quant.uniform<i8:f32, 1.0:2>>, volatile} : (tensor<1x3x1x1xf32>) -> tensor<1x3x1x1x!quant.uniform<i8:f32, 1.0:2>>
  %10 = "tfl.dequantize"(%9) : (tensor<1x3x1x1x!quant.uniform<i8:f32, 1.0:2>>) -> tensor<1x3x1x1xf32>