eazygrad.SGD¶

class eazygrad.SGD(parameters: Sequence[_Tensor], lr: float = 0.001, momentum: float = 0.0, dampening: float = 0.0)[source]¶

Bases: Optimizer

Stochastic gradient descent optimizer.

Parameters:

parameters (sequence of _Tensor) – Iterable of tensors to optimize.
lr (float, default=1e-3) – Learning rate.
momentum (float, default=0.0) – Momentum factor.
dampening (float, default=0.0) – Dampening applied to the momentum update.

Notes

If momentum is zero, the optimizer reduces to plain stochastic gradient descent. Otherwise, it maintains one momentum buffer per parameter.