struct Add: Differentiable {
    var bias: Tensor<Float>
    func applied(to input: Tensor<Float>) -> Tensor<Float> {
        return (input + bias).sum()
    }
}

let model = Add(bias: Tensor<Float>(zeros: [512]))
let x = Tensor<Float>(zeros: [4096, 512])
for i in 1...1000 {
  print(i)
  let grad = model.gradient { m in m.applied(to: x) }
}