pub struct SGD<B: Backend> {
pub learning_rate: f32,
pub momentum: f32,
pub weight_decay: f32,
pub dampening: f32,
pub nesterov: bool,
pub maximize: bool,
pub bias: Vec<Tensor<B>>,
}
Expand description
§Stochastic gradient descent optimizer
Fields§
§learning_rate: f32
learning rate (default: 0.001)
momentum: f32
momentum factor (default: 0.0)
weight_decay: f32
weight decay (L2 penalty) (default: 0.0)
dampening: f32
dampening for momentum (default: 0.0)
nesterov: bool
enables Nesterov momentum (default: false)
maximize: bool
maximize the objective with respect to the params, instead of minimizing (default: false)
bias: Vec<Tensor<B>>
stores momentum, starts empty and will be initialized on demand
Implementations§
source§impl<B: Backend> SGD<B>
impl<B: Backend> SGD<B>
sourcepub fn update<'a>(
&mut self,
parameters: impl IntoIterator<Item = &'a mut Tensor<B>>,
gradients: impl IntoIterator<Item = Option<Tensor<B>>>
)where
B: 'a,
pub fn update<'a>(
&mut self,
parameters: impl IntoIterator<Item = &'a mut Tensor<B>>,
gradients: impl IntoIterator<Item = Option<Tensor<B>>>
)where
B: 'a,
Updates parameters with gradients. Number of parameters must be the same as number of gradients. Gradients can be None, those are simply skipped.
Trait Implementations§
Auto Trait Implementations§
impl<B> Freeze for SGD<B>
impl<B> RefUnwindSafe for SGD<B>where
B: RefUnwindSafe,
impl<B> Send for SGD<B>where
B: Send,
impl<B> Sync for SGD<B>where
B: Sync,
impl<B> Unpin for SGD<B>where
B: Unpin,
impl<B> UnwindSafe for SGD<B>where
B: UnwindSafe,
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