pub enum FileType {
F32,
MostlyF16,
MostlyQ4_0,
MostlyQ4_1,
MostlyQ4_1SomeF16,
MostlyQ4_2,
MostlyQ8_0,
MostlyQ5_0,
MostlyQ5_1,
}
Expand description
How the tensors are stored in GGML LLM models.
Variants§
F32
All tensors are stored as f32.
MostlyF16
All tensors are mostly stored as f16
, except for the 1D tensors (32-bit).
MostlyQ4_0
All tensors are mostly stored as Q4_0
, except for the 1D tensors (32-bit).
MostlyQ4_1
All tensors are mostly stored as Q4_1
, except for the 1D tensors (32-bit)
MostlyQ4_1SomeF16
All tensors are mostly stored as Q4_1
, except for the 1D tensors (32-bit)
and the tok_embeddings.weight
(f16) and output.weight
tensors (f16).
MostlyQ4_2
All tensors are mostly stored as Q4_2
, except for the 1D tensors (32-bit).
MostlyQ8_0
All tensors are mostly stored as Q8_0
, except for the 1D tensors (32-bit).
MostlyQ5_0
All tensors are mostly stored as Q5_0
, except for the 1D tensors (32-bit).
MostlyQ5_1
All tensors are mostly stored as Q5_1
, except for the 1D tensors (32-bit).
Trait Implementations§
impl Copy for FileType
impl Eq for FileType
impl StructuralPartialEq for FileType
Auto Trait Implementations§
impl Freeze for FileType
impl RefUnwindSafe for FileType
impl Send for FileType
impl Sync for FileType
impl Unpin for FileType
impl UnwindSafe for FileType
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