torch_npu.optim.NpuFusedOptimizerBase

接口原型

CLASS torch_npu.optim.NpuFusedOptimizerBase(params, default)

功能描述

张量融合优化器的基类，实现梯度清零、梯度更新等优化器基本功能，用户可进行继承实现自定义融合优化器。

参数说明

输入说明

params为参数的可迭代对象或参数组的dict类型。

输出说明

类型为“NpuFusedOptimizerBase”的对象。

约束说明

NpuFusedOptimizerBase为基类，无法单独使用，需通过继承实现子类实现特定功能的融合优化器。

支持的型号

调用示例

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
import torch
import torch_npu
from torch_npu.optim.npu_fused_optim_base import NpuFusedOptimizerBase
from torch.optim.optimizer import required

LR_MIN = 0.0
MOMENTUM_MIN = 0.0
DAMPENING_DEFAULT = 0.0
WEIGHT_DECAY_MIN = 0.0


class NpuFusedSGD(NpuFusedOptimizerBase):
    def __init__(self,
                 params,
                 lr=required,
                 momentum=MOMENTUM_MIN,
                 dampening=DAMPENING_DEFAULT,
                 weight_decay=WEIGHT_DECAY_MIN,
                 nesterov=False):
        if lr is not required and lr < LR_MIN:
            raise ValueError("Invalid learning rate: {}".format(lr))
        if momentum < MOMENTUM_MIN:
            raise ValueError("Invalid momentum value: {}".format(momentum))
        if weight_decay < WEIGHT_DECAY_MIN:
            raise ValueError(
                "Invalid weight_decay value: {}".format(weight_decay))
        if nesterov and (momentum <= MOMENTUM_MIN
                         or not math.isclose(dampening, DAMPENING_DEFAULT, abs_tol=1e-15)):
            raise ValueError(
                "Nesterov momentum requires a momentum and zero dampening")
        defaults = dict(lr=lr,
                        momentum=momentum,
                        dampening=dampening,
                        weight_decay=weight_decay,
                        nesterov=nesterov)
        self._momentum_buffer_already_in_state = False
        super(NpuFusedSGD, self).__init__(params, defaults)

    def __setstate__(self, state):
        super(NpuFusedSGD, self).__setstate__(state)
        for group in self.param_groups:            
            group.setdefault('nesterov', False)
父主题： torch_npu.optim