`fluke.config`¶

`Configuration`	Fluke configuration class.
`ConfigurationError`	Exception raised when the configuration is not valid.
`OptimizerConfigurator`	This class is used to configure the optimizer and the learning rate scheduler.

class `fluke.config.Configuration`

class fluke.config.Configuration(config_exp_path: str = None, config_alg_path: str = None, force_validation: bool = True)[source]¶

Fluke configuration class. This class is used to store the configuration of an experiment. The configuration must adhere to a specific structure. The configuration is validated when the class is instantiated.

Parameters:

config_exp_path (str) – The path to the experiment configuration file.
config_alg_path (str) – The path to the algorithm configuration file.
force_validation (bool, optional) – Whether to force the validation of the configuration. Defaults to True.

Raises:

ValueError – If the configuration is not valid.

property client: DDict¶

Get quick access to the client’s hyperparameters.

Returns:: The client’s hyperparameters.
Return type:: DDict

classmethod from_dict(cfg_dict: dict | DictConfig) → Configuration[source]¶

Create a configuration from a dictionary.

Parameters:: cfg_dict (dict | DictConfig) – The dictionary.
Returns:: The configuration.
Return type:: Configuration

property model: DDict¶

Get quick access to the model hyperparameters.

Returns:: The model hyperparameters.
Return type:: DDict

property server: DDict¶

Get quick access to the server’s hyperparameters.

Returns:: The server’s hyperparameters.
Return type:: DDict

classmethod sweep(config_exp_path: str, config_alg_path: str) → list[Configuration][source]¶

Generate configurations from a sweep. This method is used to generate configurations from a sweep. The sweep is defined by the experiment configuration file. The method yields a configuration for each combination of hyperparameters.

Parameters:

config_exp_path (str) – The path to the experiment configuration file.
config_alg_path (str) – The path to the algorithm configuration file.

Returns:

A list of configurations.

Return type:

list[Configuration]

to_dict() → dict[source]¶

Convert the configuration to a dictionary.

Returns:: The dictionary.
Return type:: dict

class `fluke.config.ConfigurationError`

class fluke.config.ConfigurationError[source]¶: Exception raised when the configuration is not valid.

class `fluke.config.OptimizerConfigurator`

class fluke.config.OptimizerConfigurator(optimizer_cfg: DDict | dict, scheduler_cfg: DDict | dict = None)[source]¶

This class is used to configure the optimizer and the learning rate scheduler.

optimizer¶

The optimizer class.

Type:: type[Optimizer]

scheduler¶

The learning rate scheduler class.

Type:: type[LRScheduler]

optimizer_cfg¶

The optimizer keyword arguments.

Type:: DDict

scheduler_cfg¶

The scheduler keyword arguments.

Type:: DDict

__call__(model: Module, filter_fun: callable | None = None, **override_kwargs)[source]¶

Creates the optimizer and the scheduler.

Parameters:

model (Module) – The model whose parameters will be optimized.
filter_fun (callable) – This must be a function of the model and it must returns the set of parameters that the optimizer will consider.
override_kwargs (dict) – The optimizer’s keyword arguments to override the default ones.

Returns:

The optimizer and the scheduler.

Return type:

tuple[Optimizer, StepLR]

fluke.config¶

class fluke.config.Configuration

class fluke.config.ConfigurationError

class fluke.config.OptimizerConfigurator

`fluke.config`¶

class `fluke.config.Configuration`

class `fluke.config.ConfigurationError`

class `fluke.config.OptimizerConfigurator`