Config

class aido.config.AIDOConfig(optimizer: ~aido.config.OptimizerConfig = <factory>, surrogate: ~aido.config.SurrogateConfig = <factory>, simulation: ~aido.config.SimulationConfig = <factory>, scheduler: ~aido.config.SchedulerConfig = <factory>)

Bases: object

Config Dataclass for storing the internal parameter such as the hyperparameters of the different models and the way new values are sampled for the Simulation Task.

This class is serializable to and from json. In order to be picked up by the AIDO scheduler, the json file with updated values must be placed in the AIDO root directory.

Default fields:

• Optimizer:

  • optimizer.lr: float = 0.02 (>0)

  • optimizer.batch_size: int = 512

  • optimizer.n_epochs: int = 40

• Surrogate:

  • surrogate.n_epoch_pre: int = 24

  • surrogate.n_epochs_main: int = 40

• Simulation:

  • simulation.generate_scaling: float = 1.2 (>0)

  • simulation.sigma: float = 1.5 (>0)

  • simulation.sigma_mode: str = “flat” (or “scale”)

• Scheduler:

  • scheduler.training_num_retries: int = 20

  • scheduler.training_delay_between_retries: int | float = 60 (in seconds)

__getitem__(key: str) → Any

See AIDOConfig.get_value()

as_dict() → Dict

Return all values from this Config class as a dict

Returns:

dict – Nested dictionary with subclasses also being dicts

from_dict(new_dict: dict) → None

Update values from this instance with values from the provided dict.

Parameters:

new_dict (dict) – A dictionary with the new values, with the keys being in the dot-separated format used by AIDOConfig.set_value()

classmethod from_json(file_path: str) → Self

Create a new instance from a json file

Parameters:

file_path (str) – The input file path

Returns:

AIDOConfig – New instance of this class

Raises:

• Warning – If the file could not be found, a warning is displayed instead of raising a FileNotFoundError. This is to ensure that the code can still run despite an invalid config.

• Warning – If any Exception is raised while reading in the json file, e.g. if it is invalid or wrongly formatted, a warning is displayed. Will show the Error message for debugging.

get_key(key: str) → Tuple[Self, Any]

Helper method to find the subclass that corresponds to a given dot-separated name.

Parameters:

key (str) – A dot-separated name. For example “optimizer.lr”.

Returns:

tuple – A tuple of the sub-class together with the attribute name.

Example

>>> AIDOConfig.get_key("optimizer.lr")
(OptimizerConfig(lr=0.02, batch_size=512, n_epochs=40), 'lr')

Where the first entry is an instance of the OptimizerConfig subclass and the second entry is the key that can be used to access that attribute.

>>> getattr(OptimizerConfig, "lr")
0.02

get_value(key: str) → Any

Get the value from one of the subclasses

Parameters:

key (str) – A dot-separated name

Returns:

Any – The attribute value corresponding to the key

set_value(key: str, value: Any) → None

Change the value of a field by providing a dot-separated key and value

Parameters:

• key (str) – dot-separated name, for example “optimizer.lr” will adjust the attribute “lr” of the sub-class OptimizerConfig.

• value (Any) – The updated value

to_json(file_path: str) → None

Write the current values to a json file

Parameters:

file_path (str) – The output file path

class aido.config.OptimizerConfig(lr: float = 0.02, batch_size: int = 512, n_epochs: int = 40)

Bases: object

class aido.config.SchedulerConfig(training_num_retries: int = 20, training_delay_between_retries: int | float = 60)

Bases: object

class aido.config.SimulationConfig(generate_scaling: float = 1.2, sigma: float = 1.5, sigma_mode: str = 'flat')

Bases: object

class aido.config.SurrogateConfig(n_epoch_pre: int = 24, n_epochs_main: int = 40)

Bases: object