zzy34407230
/
mindspore2022
Not watched
1
0
Fork 0
Code
Releases 22 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
34622 Commits
18 Branches
584 MB
239 Download
Tree: 03554c47ca
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '03554c47ca'
${ noResults }
mindspore2022/docs/api/api_python/nn/mindspore.nn.Optimizer.txt

mindspore.nn.Optimizer.txt 6.1 kB
Raw Normal View History

optim loss_scale datasethelper chinese api
4 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145 
  1. Class mindspore.nn.Optimizer(learning_rate, parameters, weight_decay=0.0, loss_scale=1.0)

  2. 

  3.     ڲµŻࡣҪֱʹ࣬ʵһࡣ

  4. 

  5.     Żֲ֧顣ʱÿòͬѧϰʣ`lr`Ȩ˥`weight_decay`ݶĻ`grad_centralization`ԡ

  6. 

  7.     ע

  8.         ڲδʱŻõ`weight_decay`Ӧƺ"beta""gamma"ͨɵȨ˥ԡʱÿ`weight_decay`δãʹŻõ`weight_decay`

  9. 

  10.     

  11.         learning_rate (Union[float, int, Tensor, Iterable, LearningRateSchedule]):

  12. 

  13.             - float: ̶ѧϰʡڵ㡣

  14. 

  15.             - int: ̶ѧϰʡڵ㡣ͻᱻתΪ

  16. 

  17.             - Tensor: DZһάǹ̶ѧϰʡһάǶ̬ѧϰʣiȡеiֵΪѧϰʡ

  18. 

  19.             - Iterable: ̬ѧϰʡiȡiֵΪѧϰʡ

  20. 

  21.             - LearningRateSchedule: ̬ѧϰʡѵУŻʹòstepΪ룬`LearningRateSchedule`ʵ㵱ǰѧϰʡ

  22. 

  23.         parameters (Union[list[Parameter], list[dict]]): `Parameter`ɵбֵɵббԪֵʱֵļ"params""lr""weight_decay""grad_centralization""order_params"

  24. 

  25.             - params: ǰȨأֵ`Parameter`б

  26. 

  27.             - lr: ѡд"lr"ʹöӦֵΪѧϰʡ

  28.               ûУʹŻõ`learning_rate`Ϊѧϰʡ

  29. 

  30.             - weight_decay: ѡд"weight_decayʹöӦֵΪȨ˥ֵûУʹŻõ`weight_decay`ΪȨ˥ֵ

  31. 

  32.             - grad_centralization: ѡд"grad_centralization"ʹöӦֵֵΪ͡ûУΪ`grad_centralization`ΪFalse

  33.               òھ㡣

  34. 

  35.             - order_params: ѡӦֵԤڵIJ˳򡣵ʹò鹦ʱͨʹø`parameters`˳ܡ

  36.               д"order_params"Ըе"order_params"еIJijһ`params`С

  37. 

  38.         weight_decay (Union[float, int]): Ȩ˥򸡵ֵ

  39.             ڻ0

  40.             `weight_decay`תΪĬֵ0.0

  41. 

  42.         loss_scale (float): ݶϵ0`loss_scale`תΪͨʹĬֵѵʱʹ`FixedLossScaleManager``FixedLossScaleManager``drop_overflow_update`ΪFalseʱֵҪ`FixedLossScaleManager`е`loss_scale`ͬйظϸϢclass`mindspore.FixedLossScaleManager`

  43.             Ĭֵ1.0

  44. 

  45.     쳣

  46.         TypeError: `learning_rate`intfloatTensorIterableLearningRateSchedule

  47.         TypeError: `parameters`ԪزParameterֵ䡣

  48.         TypeError: `loss_scale`float

  49.         TypeError: `weight_decay`floatint

  50.         ValueError: `loss_scale`Сڻ0

  51.         ValueError: `weight_decay`С0

  52.         ValueError: `learning_rate`һTensorTensorάȴ1

  53. 

  54.     ֵ֧ƽ̨

  55.         ``Ascend`` ``GPU`` ``CPU``

  56. 

  57. 

  58. broadcast_params(optim_result)

  59. 

  60.         ˳в㲥

  61. 

  62.         

  63.             optim_result (bool): ½֤ɺִв㲥

  64. 

  65.         أ

  66.              bool״̬־

  67. 

  68. 

  69. decay_weight(gradients)

  70. 

  71.         ˥Ȩء

  72. 

  73.         һּѧϰģ͹ϵķ̳ :class:`mindspore.nn.Optimizer`ԶŻʱɵøýӿڽȨ˥

  74. 

  75.         

  76.             gradients (tuple[Tensor]): ݶȣ״shapeͬ

  77. 

  78.         أ

  79.             tuple[Tensor]˥Ȩغݶȡ

  80. 

  81. 

  82. get_lr()

  83. 

  84.         Żøýӿڻȡǰ裨stepѧϰʡ̳ :class:`mindspore.nn.Optimizer`ԶŻʱڲǰøýӿڻȡѧϰʡ

  85. 

  86.         أ

  87.             floatǰѧϰʡ

  88. 

  89. 

  90. get_lr_parameter(param)

  91. 

  92.         ʹ鹦ܣΪͬòͬѧϰʱȡָѧϰʡ

  93. 

  94.         

  95.             param (Union[Parameter, list[Parameter]]): `Parameter``Parameter`б

  96. 

  97.         أ

  98.             Parameter`Parameter``Parameter`бʹ˶̬ѧϰʣڼѧϰʵ`LearningRateSchedule``LearningRateSchedule`б

  99. 

  100.         ʾ

  101.             >>> from mindspore import nn

  102.             >>> net = Net()

  103.             >>> conv_params = list(filter(lambda x: 'conv' in x.name, net.trainable_params()))

  104.             >>> no_conv_params = list(filter(lambda x: 'conv' not in x.name, net.trainable_params()))

  105.             >>> group_params = [{'params': conv_params, 'lr': 0.05},

  106.             ...                 {'params': no_conv_params, 'lr': 0.01}]

  107.             >>> optim = nn.Momentum(group_params, learning_rate=0.1, momentum=0.9, weight_decay=0.0)

  108.             >>> conv_lr = optim.get_lr_parameter(conv_params)

  109.             >>> print(conv_lr[0].asnumpy())

  110.             0.05

  111. 

  112. 

  113. gradients_centralization(gradients)

  114. 

  115.         ݶĻ

  116. 

  117.         һŻѧϰģѵٶȵķ̳ :class:`mindspore.nn.Optimizer`ԶŻʱɵøýӿڽݶĻ

  118. 

  119.         

  120.             gradients (tuple[Tensor]): ݶȣ״shapeͬ

  121. 

  122.         أ

  123.             tuple[Tensor]ݶĻݶȡ

  124. 

  125. 

  126. scale_grad(gradients)

  127. 

  128.         ڻϾȳԭݶȡ

  129. 

  130.         ̳:class:`mindspore.nn.Optimizer`ԶŻʱɵøýӿڻԭݶȡ

  131. 

  132.         

  133.             gradients (tuple[Tensor]): ݶȣ״shapeͬ

  134. 

  135.         أ

  136.             tuple[Tensor]ԭݶȡ

  137. 

  138. 

  139. target

  140. 

  141.         ָhostϻ豸deviceϸ²Ϊstrֻ'CPU''Ascend''GPU'

  142. 

  143. 

  144. unique

  145.         ԱʾǷŻнݶȥأͨϡ硣ݶϡΪTrueǰϡѶȨȥأݶdzܵģΪFalseδʱĬֵΪTrue

阿对对队