3.1.10. aijack.collaborative.optimizer package#

3.1.10.1. Submodules#

3.1.10.2. aijack.collaborative.optimizer.adam module#

class aijack.collaborative.optimizer.adam.AdamFLOptimizer(parameters, lr=0.01, weight_decay=0.0001, beta1=0.9, beta2=0.999, epsilon=1e-08)[source]#

Bases: aijack.collaborative.optimizer.base.BaseFLOptimizer

Implementation of Adam to update the global model of Federated Learning.

Parameters

  • parameters (List[torch.nn.Parameter]) – parameters of the model

  • lr (float, optional) – learning rate. Defaults to 0.01.

  • weight_decay (float, optional) – coefficient of weight decay. Defaults to 0.0001.

  • beta1 (float, optional) – 1st-order exponential decay. Defaults to 0.9.

  • beta2 (float, optional) – 2nd-order exponential decay. Defaults to 0.999.

  • epsilon (float, optional) – a small value to prevent zero-devision. Defaults to 1e-8.

step(grads)[source]#

Update the parameters with the give gradient

Parameters

grads (List[torch.Tensor]) – list of gradients

3.1.10.3. aijack.collaborative.optimizer.base module#

class aijack.collaborative.optimizer.base.BaseFLOptimizer(parameters, lr=0.01, weight_decay=0.0001)[source]#

Bases: object

Basic class for optimizers of the server in Federated Learning.

Parameters

  • parameters (List[torch.nn.Parameter]) – parameters of the model

  • lr (float, optional) – learning rate. Defaults to 0.01.

  • weight_decay (float, optional) – coefficient of weight decay. Defaults to 0.0001.

abstract step(grads)[source]#

Update the parameters with the give gradient

Parameters

grads (List[torch.Tensor]) – list of gradients

3.1.10.4. aijack.collaborative.optimizer.sgd module#

class aijack.collaborative.optimizer.sgd.SGDFLOptimizer(parameters, lr=0.01, weight_decay=0.0)[source]#

Bases: aijack.collaborative.optimizer.base.BaseFLOptimizer

Implementation of SGD to update the global model of Federated Learning.

Parameters

  • parameters (List[torch.nn.Parameter]) – parameters of the model

  • lr (float, optional) – learning rate. Defaults to 0.01.

  • weight_decay (float, optional) – coefficient of weight decay. Defaults to 0.0001.

step(grads)[source]#

Update the parameters with the give gradient

Parameters

grads (List[torch.Tensor]) – list of gradients

3.1.10.5. Module contents#

Implementation of basic collaborative optimizers for neural network

class aijack.collaborative.optimizer.AdamFLOptimizer(parameters, lr=0.01, weight_decay=0.0001, beta1=0.9, beta2=0.999, epsilon=1e-08)[source]#

Bases: aijack.collaborative.optimizer.base.BaseFLOptimizer

Implementation of Adam to update the global model of Federated Learning.

Parameters

  • parameters (List[torch.nn.Parameter]) – parameters of the model

  • lr (float, optional) – learning rate. Defaults to 0.01.

  • weight_decay (float, optional) – coefficient of weight decay. Defaults to 0.0001.

  • beta1 (float, optional) – 1st-order exponential decay. Defaults to 0.9.

  • beta2 (float, optional) – 2nd-order exponential decay. Defaults to 0.999.

  • epsilon (float, optional) – a small value to prevent zero-devision. Defaults to 1e-8.

step(grads)[source]#

Update the parameters with the give gradient

Parameters

grads (List[torch.Tensor]) – list of gradients

class aijack.collaborative.optimizer.SGDFLOptimizer(parameters, lr=0.01, weight_decay=0.0)[source]#

Bases: aijack.collaborative.optimizer.base.BaseFLOptimizer

Implementation of SGD to update the global model of Federated Learning.

Parameters

  • parameters (List[torch.nn.Parameter]) – parameters of the model

  • lr (float, optional) – learning rate. Defaults to 0.01.

  • weight_decay (float, optional) – coefficient of weight decay. Defaults to 0.0001.

step(grads)[source]#

Update the parameters with the give gradient

Parameters

grads (List[torch.Tensor]) – list of gradients