autograd 1.1.1

Tensors and differentiable operations in Rust
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

use crate::Float;

pub(crate) struct SGDOp<T: Float> {
    pub lr: T,
}

impl<T: Float> SGDOp<T> {
    pub(crate) fn new(lr: T) -> Self {
        SGDOp { lr }
    }
}

impl<T: Float> crate::op::Op<T> for SGDOp<T> {
    fn compute(&self, ctx: &mut crate::op::ComputeContext<T>) {
        ctx.input_mut(0).scaled_add(-self.lr, &ctx.input(1));
        ctx.append_empty_output();
    }

    fn grad(&self, ctx: &mut crate::op::GradientContext<T>) {
        ctx.append_input_grad(None);
    }
}