pub enum RoutedExpertTensorParallelOutput<T> {
Complete(T),
Partial(TensorParallelGroupedOutput<T>),
}Expand description
Provider result that distinguishes complete outputs from rank-local TP work.
Variants§
Complete(T)
Provider already completed every required collective and bias addition.
Partial(TensorParallelGroupedOutput<T>)
Caller must all-sum reducible, then add post_reduce exactly once.
Auto Trait Implementations§
impl<T> Freeze for RoutedExpertTensorParallelOutput<T>
impl<T> RefUnwindSafe for RoutedExpertTensorParallelOutput<T>
impl<T> Send for RoutedExpertTensorParallelOutput<T>
impl<T> Sync for RoutedExpertTensorParallelOutput<T>
impl<T> Unpin for RoutedExpertTensorParallelOutput<T>
impl<T> UnsafeUnpin for RoutedExpertTensorParallelOutput<T>
impl<T> UnwindSafe for RoutedExpertTensorParallelOutput<T>
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