Enum ParityStatus 

pub enum ParityStatus {
    Verified {
        fixture: &'static str,
    },
    Unverified,
}

Whether this model’s numerics have been checked against a Python reference.

Unverified is the honest default and covers most rows. Read it as “this model’s output could be anything”, not as “this model is probably fine”.

That wording used to be softer — it said Unverified was not a statement that a model is wrong, only that nothing proved it right. ProteinMPNN then demonstrated the difference is not academic: it sat at Unverified through every release up to v0.3.3 while agreeing with the reference on 2 of 93 positions, where chance alone over a 21-token vocabulary is about 4. The output was not approximately right, it was unrelated to what the model computes, and the registry said only “unchecked” the whole time (ferritin-100.33).

The lesson is about which rows are dangerous. A row whose family already has a Verified sibling shares a proven code path and differs mainly in weights. A row whose family has no verified member has an entire architecture that nothing has ever checked — that is the position ProteinMPNN was in. Prefer closing whole-family gaps over adding a second fixture to a family that already has one.

Variants

Verified

A committed fixture pins this model’s output against the reference implementation. Names the fixture stem under tests/fixtures/.

Fields

fixture: &'static str

Fixture stem, without the .safetensors extension.

Unverified

No parity fixture exists, so nothing checks this model’s numerics.

Not a weaker form of Verified — an absence of evidence that has already, once, been concealing a model that was flatly wrong.

Trait Implementations

impl Clone for ParityStatus

fn clone(&self) -> ParityStatus

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for ParityStatus

impl Debug for ParityStatus

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for ParityStatus

impl PartialEq for ParityStatus

fn eq(&self, other: &ParityStatus) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for ParityStatus