Enum GgufValue 

pub enum GgufValue {
    U8(u8),
    I8(i8),
    U16(u16),
    I16(i16),
    U32(u32),
    I32(i32),
    U64(u64),
    I64(i64),
    F32(f32),
    F64(f64),
    Bool(bool),
    String(String),
    Array(Vec<GgufValue>),
}

A single metadata value, typed as GGUF’s key-value store types it.

Modeling this as a strongly-typed enum — rather than handing callers the raw bytes and a type tag to match on themselves — means “get me llama.block_count as a u64” is one call (GgufMetadata::get_u64) instead of a match arm duplicated at every call site.

Variants

U8(u8)

I8(i8)

U16(u16)

I16(i16)

U32(u32)

I32(i32)

U64(u64)

I64(i64)

F32(f32)

F64(f64)

Bool(bool)

String(String)

Array(Vec<GgufValue>)

A (possibly nested) array of values. GGUF arrays are homogeneous in principle, but the format does not forbid nesting Array inside Array, so this has to allow it too.

Implementations

impl GgufValue

pub fn as_u64(&self) -> Option<u64>

Widens to u64 for every unsigned integer variant, None otherwise.

The GGUF spec changed most count/length fields from uint32 to uint64 between v1 and v2 and explicitly recommends readers accept both (“Some models may use uint32 for their values; it is recommended that readers support both”). Widening here, once, is how this crate honors that recommendation without every call site re-deriving it.

pub fn as_u32(&self) -> Option<u32>

Narrows to u32, refusing to silently truncate a u64 that does not fit — a truncated block_count is a corrupt-looking model, not a smaller one, so this returns None rather than lying.

pub fn as_i64(&self) -> Option<i64>

Widens to i64 for every signed integer variant, None otherwise.

pub fn as_i32(&self) -> Option<i32>

Narrows to i32, refusing to silently truncate.

pub fn as_f32(&self) -> Option<f32>

The exact f32, if this value was stored as one. Deliberately does not narrow F64 to F32: unlike the integer widenings above, float narrowing is lossy in the common case, not just at the edges, so silently doing it would be more surprising than useful.

pub fn as_f64(&self) -> Option<f64>

Widens to f64 for either float variant.

pub fn as_bool(&self) -> Option<bool>

pub fn as_str(&self) -> Option<&str>

pub fn as_array(&self) -> Option<&[GgufValue]>

Trait Implementations

impl Clone for GgufValue

fn clone(&self) -> GgufValue

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GgufValue

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

Formats the value using the given formatter.

impl PartialEq for GgufValue

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

Equality operator ==.

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

Inequality operator !=.

impl StructuralPartialEq for GgufValue