pub fn induce_exact(
provider: &CudaKernelProvider,
request: &InduceExactRequest<'_>,
) -> Result<ExactInductionResult>Expand description
Run exact induction against one request.
Returns an ExactInductionResult matching the Python reference on
bounded inputs.
The provider argument owns the kernel launcher — it’s passed separately
from the request so the engine can also materialize short-lived GPU
buffers (an empty negatives buffer when request.negatives is None).