forked from OSSInnovation/mindspore
add optimizer formula and verification
This commit is contained in:
parent
12a150bb5d
commit
dc7cc66b3d
|
@ -53,6 +53,19 @@ class Momentum(Optimizer):
|
|||
|
||||
To improve parameter groups performance, the customized order of parameters can be supported.
|
||||
|
||||
.. math::
|
||||
v_{t} = v_{t-1} \ast u + gradients
|
||||
|
||||
If use_nesterov is True:
|
||||
.. math::
|
||||
p_{t} = grad \ast lr + v_{t} \ast u \ast lr
|
||||
|
||||
If use_nesterov is Flase:
|
||||
.. math::
|
||||
p_{t} = lr \ast v_{t}
|
||||
|
||||
Here: where grad, lr, p, v and u denote the gradients, learning_rate, parameter, accum, and momentum respectively.
|
||||
|
||||
Args:
|
||||
params (Union[list[Parameter], list[dict]]): When the `params` is a list of `Parameter` which will be updated,
|
||||
the element in `params` should be class `Parameter`. When the `params` is a list of `dict`, the "params",
|
||||
|
|
|
@ -46,6 +46,21 @@ class SGD(Optimizer):
|
|||
|
||||
To improve parameter groups performance, the customized order of parameters can be supported.
|
||||
|
||||
.. math::
|
||||
v_{t+1} = u \ast v_{t} + gradient \ast (1-dampening)
|
||||
|
||||
If nesterov is True:
|
||||
.. math::
|
||||
p_{t+1} = p_{t} - lr \ast (gradient + u \ast v_{t+1})
|
||||
|
||||
If nesterov is Flase:
|
||||
.. math::
|
||||
p_{t+1} = p_{t} - lr \ast v_{t+1}
|
||||
|
||||
To be notice, for the first step, v_{t+1} = gradient
|
||||
|
||||
Here : where p, v and u denote the parameters, accum, and momentum respectively.
|
||||
|
||||
Args:
|
||||
params (Union[list[Parameter], list[dict]]): When the `params` is a list of `Parameter` which will be updated,
|
||||
the element in `params` should be class `Parameter`. When the `params` is a list of `dict`, the "params",
|
||||
|
@ -74,7 +89,8 @@ class SGD(Optimizer):
|
|||
momentum (float): A floating point value the momentum. should be at least 0.0. Default: 0.0.
|
||||
dampening (float): A floating point value of dampening for momentum. should be at least 0.0. Default: 0.0.
|
||||
weight_decay (float): Weight decay (L2 penalty). It should be in range [0.0, 1.0]. Default: 0.0.
|
||||
nesterov (bool): Enables the Nesterov momentum. Default: False.
|
||||
nesterov (bool): Enables the Nesterov momentum. If use nesterov, momentum must greater then 0,
|
||||
and dampening must equal to 1. Default: False.
|
||||
loss_scale (float): A floating point value for the loss scale. Should be not less than 1.0. Default: 1.0.
|
||||
|
||||
Inputs:
|
||||
|
@ -118,6 +134,10 @@ class SGD(Optimizer):
|
|||
if isinstance(momentum, float) and momentum < 0.0:
|
||||
raise ValueError("momentum should be at least 0.0, but got momentum {}".format(momentum))
|
||||
|
||||
if nesterov and (momentum <= 0 or dampening != 0):
|
||||
raise ValueError("If use nesterov, momentum must be positive and dampening must equal to 0,"
|
||||
"but got momentum {}, dampening {}".format(momentum, dampening))
|
||||
|
||||
if isinstance(dampening, int):
|
||||
dampening = float(dampening)
|
||||
if not isinstance(dampening, float):
|
||||
|
|
Loading…
Reference in New Issue