fix Constant Activity in examples to Const

[kiranshila]

No description provided.

[kiranshila]

agh my formatter seems to have mucked with a few things

[kiranshila]

Also, not sure where to ask about this - but in the examples, you have the likes of

#[autodiff(backprop, Reverse, Const, Duplicated, Active)]
fn training_loss(images: &[f32], weights: &[f32]) -> f32 {
  let loss = do_some_math(images, weights);
  loss
}

generating code like

fn backprop(images: &[f32], weights: &[f32], dweights: &mut [f32]) -> f32

but this isn't quite correct, right? There's an extra f32 seed parameter, if I'm not mistaken.

[ZuseZ4]

Thanks!
Yes, some parts were unfortunately written before we had CI working, so we forgot to update them when changing autodiff. I'm currently in the process of upstreaming all tests into rust/tests, and in the future let the Rust CI make sure that we don't regress them again. I also have more real-world examples these days of differentiating e.g. Karpathy's llm.c (a rust version of it), as well as Microsoft's ADBench implementation, so I intend to completely replace the outdated examples at some point.