pub struct MulMvExtParams {
pub m: u32,
pub n: u32,
pub k: u32,
pub batch: u32,
pub ggml_type: GgmlType,
}Expand description
Host-side parameters for mul_mv_ext_dispatch.
Buffer layout contract:
weight(= src0): row-major[N, blocks_per_row]GGUF blocks.input(= src1): row-major[batch, M, K]f32. K = ne00.output(= dst): row-major[batch, M, N]f32. N = ne01 / ne0.
r2, r3 model llama.cpp’s batch-broadcast (default 1, 1).
Fields§
§m: u32M — number of src1 rows (small batch, must be ∈ [2, 8]).
n: u32N — number of weight rows (output dim).
k: u32K — contract dim (input dim, must be divisible by 32).
batch: u32Batch-broadcast factor for src0 vs src1 (typical 1).
ggml_type: GgmlTypeGGUF weight type. Phase 1 supports Q5_1 + IQ4_NL only; other types
return MlxError::InvalidArgument.
Trait Implementations§
Source§impl Clone for MulMvExtParams
impl Clone for MulMvExtParams
Source§fn clone(&self) -> MulMvExtParams
fn clone(&self) -> MulMvExtParams
Returns a duplicate of the value. Read more
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moreSource§impl Debug for MulMvExtParams
impl Debug for MulMvExtParams
impl Copy for MulMvExtParams
Auto Trait Implementations§
impl Freeze for MulMvExtParams
impl RefUnwindSafe for MulMvExtParams
impl Send for MulMvExtParams
impl Sync for MulMvExtParams
impl Unpin for MulMvExtParams
impl UnsafeUnpin for MulMvExtParams
impl UnwindSafe for MulMvExtParams
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more