// CHECK-NEXT: return %[[OUTPUT_STATE]], %[[OUTPUT]] : tensor<2xui64>, tensor<10x12xui32>
  %output_key, %output = "tf.XlaRngBitGenerator"(%cst_0, %arg0, %cst) : (tensor<i32>, tensor<2xui64>, tensor<2xi32>) -> (tensor<2xui64>, tensor<10x12xui32>)
  func.return %output_key, %output : tensor<2xui64>, tensor<10x12xui32>
}

  %1 = "tf.MatMul"(%arg0, %arg0) {device = "", transpose_a = false, transpose_b = false} : (tensor<10x10xi32>, tensor<10x10xi32>) -> tensor<10x10xi32>
  %2 = "tf.PreventGradient"(%0) : (tensor<10x10xi32>) -> tensor<10x10xi32>
  %3 = "tf.PreventGradient"(%1) : (tensor<10x10xi32>) -> tensor<10x10xi32>
  %4 = "tf.AddV2"(%2, %3) {device = ""} : (tensor<10x10xi32>, tensor<10x10xi32>) -> tensor<10x10xi32>

    return %values : tensor<10x10xbf16>
  }

  // CHECK-LABEL: fusedBatchNormV3_noTraining

// CHECK-LABEL:   func @not_convert_remainder_for_uint32(
// CHECK-NOT:       "tf.Mod"
func.func @not_convert_remainder_for_uint32(%arg0: tensor<10x8xui32>, %arg1: tensor<10x8xui32>) -> tensor<10x8xui32> {
  %0 = mhlo.remainder %arg0, %arg1 : tensor<10x8xui32>
  func.return %0 : tensor<10x8xui32>
}

// CHECK-LABEL:   func @not_convert_remainder_for_uint64(
// CHECK-NOT:       "tf.Mod"