pub fn plan_depth_prior(
archive: &PriorArchive,
vocab_path: impl AsRef<Path>,
expected_geometry: &TransformerGeometry,
seed: u64,
expected_target_shapes: &BTreeMap<String, Vec<usize>>,
) -> Result<PriorPlan, PriorError>Expand description
Validates a loaded QPI1 prior archive against the target model geometry, vocabulary, and seed.
Mirrors load_depth_prior validation in nn.py without loading tensors into an ML backend.
ยงErrors
Returns PriorError if any magic, geometry, vocab SHA-256, seed, tensor set, or shape
verification fails.