// Float16-DAG: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<16xf16>
// Float16-DAG: %[[const:.*]] = "tfl.no_value"() <{value}> : () -> none
// Float16-DAG: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<3x3x3x8x16xf16>) -> tensor<3x3x3x8x16xf32>
// Float16-DAG: %[[dq_b:.*]] = "tfl.dequantize"(%[[b]]) : (tensor<16xf16>) -> tensor<16xf32>

	{TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA256, 16, 32, 16, ecdheRSAKA, suiteECDHE | suiteTLS12, cipherAES, macSHA256, nil},
	{TLS_ECDHE_RSA_WITH_AES_128_CBC_SHA, 16, 20, 16, ecdheRSAKA, suiteECDHE, cipherAES, macSHA1, nil},
	{TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA256, 16, 32, 16, ecdheECDSAKA, suiteECDHE | suiteECSign | suiteTLS12, cipherAES, macSHA256, nil},

      (tensor<4x256xf16>, tensor<256xf16>, tensor<256xf16>, tensor<256xf16>,
        tensor<256xf16>) -> tensor<4x256xf16>
  func.return %0 : tensor<4x256xf16>
}

// Validate that epsilon is overflow
func.func @batchNormInference_f16_overflow(
    %x: tensor<4x256xf16>, %scale: tensor<256xf16>, %offset: tensor<256xf16>,
    %mean: tensor<256xf16>, %variance: tensor<256xf16>)
    -> (tensor<4x256xf16>) {

  %2 = func.call @fold_all_test(%arg0, %0, %1) : (tensor<256x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<256x30x30x16xf32>
  func.return %2 : tensor<256x30x30x16xf32>
}

// ALL-LABEL: @fold_all_test
func.func @fold_all_test(%arg0: tensor<256x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>) -> tensor<256x30x30x16xf32> {

  %min = arith.constant dense<0.0> : tensor<16xf32>
  %max = arith.constant dense<255.0> : tensor<16xf32>
  %mini = "tf.Identity"(%min) : (tensor<16xf32>) -> tensor<16xf32>
  %maxi = "tf.Identity"(%max) : (tensor<16xf32>) -> tensor<16xf32>
  %fq = "tf.FakeQuantWithMinMaxVarsPerChannel"(%in, %mini, %maxi) {num_bits = 5, narrow_range = false} : (tensor<3x3x3x16xf32>, tensor<16xf32>, tensor<16xf32>) -> tensor<3x3x3x16xf32>

  %min = arith.constant dense<0.0> : tensor<16xf32>
  %max = arith.constant dense<15.0> : tensor<16xf32>
  %mini = "tf.Identity"(%min) : (tensor<16xf32>) -> tensor<16xf32>
  %maxi = "tf.Identity"(%max) : (tensor<16xf32>) -> tensor<16xf32>
  %fq = "tf.FakeQuantWithMinMaxVarsPerChannel"(%in, %mini, %maxi) {num_bits = 3, narrow_range = false} : (tensor<3x3x3x16xf32>, tensor<16xf32>, tensor<16xf32>) -> tensor<3x3x3x16xf32>

  %1 = "tfl.pseudo_const"() { value = dense<1.0> : tensor<16xf32>} : () -> tensor<16xf32>
  %2 = "tfl.conv_3d"(%arg0, %0, %1) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, dilation_d_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32, stride_d = 1 : i32} : (tensor<1x32x32x32x8xf32>, tensor<1x1x1x8x16x!quant.uniform<i8:f32, 1.0>>, tensor<16xf32>) -> tensor<1x32x32x32x16xf32>

  pass_config.unfold_batch_matmul = toco_flags.unfold_batchmatmul();
  pass_config.lower_tensor_list_ops = toco_flags.lower_tensor_list_ops();
  // Disable the unfolding of the 16x16 TF::BatchMatMulOp to avoid the
  // conversion to an unsupported 16x16 TFL::FullyConnectedOp.
  if (toco_flags.inference_type() == toco::IODataType::QUANTIZED_INT16) {
    pass_config.unfold_batch_matmul = false;
  }

}

// CHECK-LABEL: func @testLeakyRelu
func.func @testLeakyRelu(%arg0 : tensor<16xf32>) -> (tensor<16xf32>, tensor<f32>, tensor<f32>, tensor<16xf32>) {
  %pos = arith.constant dense<5.0> : tensor<f32>
  %neg = arith.constant dense<-5.0> : tensor<f32>
  %no = "tf.LeakyRelu"(%arg0) {alpha = 0.2 : f32} : (tensor<16xf32>) -> tensor<16xf32>
  %0 = "tf.LeakyRelu"(%pos) {alpha = 0.3 : f32} : (tensor<f32>) -> tensor<f32>

  %b = arith.constant dense<0.0> : tensor<16xf32>
  %conv_3d = "tfl.conv_3d"(%arg0, %w, %b) {dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x32x32x32x8xf32>, tensor<1x1x1x8x16xf32>, tensor<16xf32>) -> tensor<1x32x32x32x16xf32>
  func.return %conv_3d : tensor<1x32x32x32x16xf32>