Enum QuantScheme 

pub enum QuantScheme {
    AffineGroupInt,
    BlockScaledFloat,
    KQuantMixed,
    RtnBlock,
    Unquantized,
    Unknown,
}

The numeric format a checkpoint’s weights are stored in.

Deliberately says nothing about which engine serves the file — ModelSource already carries that, and keying this on the container instead of the format produces falsehoods: whisper.cpp’s q5_0 ggml checkpoints use the same round-to-nearest block format as llama.cpp’s, and a Gguf* variant would assert a GGUF text path that cannot load them.

What it does express is the axis neither the bit width nor the source can: an MLX 4-bit affine checkpoint and a Q4_K_M are both “4-bit”, were produced by different algorithms, and do not have the same quality.

Variants

AffineGroupInt

Integer weights with a scale and bias shared across group_size elements (4bit, 6bit). MLX’s default and only affine format.

BlockScaledFloat

Block-scaled float — MX formats (mxfp4, mxfp8). A genuinely different loader path from QuantScheme::AffineGroupInt at the same nominal width; see the microscaling-mode rejection in backend/mlx.rs, which currently accepts exactly one of these.

KQuantMixed

Mixed per-tensor bit allocation, optionally importance-weighted (Q4_K_M, Q5_K_S, IQ4_XS, TQ1_0).

RtnBlock

Uniform round-to-nearest blocks, no per-tensor mixing and no importance weighting (Q8_0, q5_0, Q4_0_4_4).

Unquantized

Full-precision weights (bf16, f16, f32). Distinct from None at the ModelSchema::quantization level: a positive claim that the checkpoint is unquantized, not an absence of information.

Unknown

A format this parser could not identify. The label is still preserved verbatim; only the classification is missing.

Trait Implementations

impl Clone for QuantScheme

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for QuantScheme

impl Debug for QuantScheme

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

Formats the value using the given formatter.

impl Default for QuantScheme

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for QuantScheme

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for QuantScheme

impl Hash for QuantScheme

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl JsonSchema for QuantScheme

fn schema_name() -> String

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn is_referenceable() -> bool

Whether JSON Schemas generated for this type should be re-used where possible using the $ref keyword.

impl PartialEq for QuantScheme

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for QuantScheme

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for QuantScheme