func.func @testBatchToSpaceInvalidOutputBatch(%arg0: tensor<16x8x8x3xf32>, %arg1: tensor<*xi32>) {
  // expected-error @+1 {{'tf.BatchToSpace' op requires output batch (dimension 0) to be equal to input batch (dimension 0) / (block_size * block_size), but got output batch 8, input batch 16, and block_size 2}}
  %0 = "tf.BatchToSpace"(%arg0, %arg1) {block_size = 2 : i64} : (tensor<16x8x8x3xf32>, tensor<*xi32>) -> tensor<8x8x8x3xf32>
  func.return
}

// -----

  func.return %0 : tensor<16x8x8x1xf32>
  }
}

// CHECK: module {
// CHECK-NEXT:   func.func @main(%arg0: tensor<8x8x1x207xf32>, %arg1: tensor<3x3x16x207xf32>) -> tensor<16x8x8x1xf32> {

func.func @testConv(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>) -> tensor<256x30x30x16xf32> {
  // CHECK: tac.device = "GPU", tac.inference_type = "FLOAT"
  %0 = "tfl.conv_2d"(%arg0, %arg1, %arg2) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<256x30x30x16xf32>

  %0 = "tfl.pseudo_const"() {value = dense<1.000000e+00> : tensor<16x3x3x3xf32>} : () -> tensor<16x3x3x3xf32>
  %1 = "tfl.pseudo_const"() {value = dense<1.000000e+00> : tensor<16xf32>} : () -> tensor<16xf32>
  %2 = func.call @fold_all_test(%arg0, %0, %1) : (tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<256x30x30x16xf32>
  func.return %2 : tensor<256x30x30x16xf32>
}

  func.return %0 : tensor<256x32x32x16xf32>
}

func.func @testConv2DGPU(tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<256x32x32x16xf32> {
^bb0(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>):
  // CHECK: tac.cost = 0x4C300000

  func.return %0 : tensor<256x32x32x16xf32>
}

func.func @testConv2DDynamicShape(tensor<?x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<?x32x32x16xf32> {
^bb0(%arg0: tensor<?x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>):
  // CHECK: _arithmetic_count = -1 : i64

// CHECK: return %[[RES0]] : tensor<256x32x32x16xf32>

{
  "version": 3,
  "operator_codes": [
    {
      "builtin_code": "CONV_2D"
    }
  ],
  "subgraphs": [
    {

// CHECK: return %[[RES0]] : tensor<256x32x32x16xf32>

{
  "version": 3,
  "operator_codes": [
    {
      "builtin_code": "CONV_2D"
    }
  ],
  "subgraphs": [

// permutated to match the new input shape.
// CHECK: (tensor<1x16x16x4xf32>, tensor<f32>) -> tensor<1x8x8x4xf32>

// Check that a `stablehlo.transpose` is added to the result to match the shape
// of the users.
// CHECK: %[[TRANSPOSE:.+]] = stablehlo.transpose %[[REDUCE_WINDOW]], dims = [0, 3, 1, 2] : (tensor<1x8x8x4xf32>) -> tensor<1x4x8x8xf32>
// CHECK: return %[[TRANSPOSE]]

// -----

func.func @fusedConv2dRelu(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>) -> tensor<256x32x32x16xf32> {
  %0 = "tfl.conv_2d"(%arg0, %arg1, %arg2) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<256x32x32x16xf32>