Skip to main content

run_token

Function run_token 

pub fn run_token<H, K>(
    plan: &DensePlan<'_, H>,
    kernels: &mut K,
    workspace: DenseWorkspace<'_>,
    token: u32,
    position: usize,
    byte_budget: usize,
) -> Result<DenseStep, DenseError<<K as TensorKernels<H>>::Error>>
where H: Copy, K: TensorKernels<H>,
Expand description

Executes a complete dense token and finite greedy argmax; failure can leave partial KV/scratch updates. Caller-owned logits remain available for inspection.