Function benchmark_forward_backward 

pub fn benchmark_forward_backward<B, M, F, L>(
    model: &M,
    input_fn: F,
    loss_fn: L,
    batch_size: usize,
    warmup: usize,
    iterations: usize,
) -> Result<BenchmarkResult>
where
    B: Backend,
    M: Module<B>,
    F: Fn() -> Tensor<B>,
    L: Fn(&Tensor<B>) -> Result<Tensor<B>>,

Benchmark forward + backward pass.

Same as benchmark_forward but also runs backpropagation on each iteration.