advsecurenet.shared.types.configs package

---

• advsecurenet.shared.types.configs.attack_configs package
  • advsecurenet.shared.types.configs.attack_configs.attack_config module
    • AttackConfig
      • AttackConfig.device
      • AttackConfig.targeted
  • advsecurenet.shared.types.configs.attack_configs.attacker_config module
    • AttackerConfig
      • AttackerConfig.attack
      • AttackerConfig.dataloader
      • AttackerConfig.device
      • AttackerConfig.evaluators
      • AttackerConfig.model
      • AttackerConfig.return_adversarial_images
  • advsecurenet.shared.types.configs.attack_configs.cw_attack_config module
    • CWAttackConfig
      • CWAttackConfig.abort_early
      • CWAttackConfig.binary_search_steps
      • CWAttackConfig.c_init
      • CWAttackConfig.c_lower
      • CWAttackConfig.c_upper
      • CWAttackConfig.clip_max
      • CWAttackConfig.clip_min
      • CWAttackConfig.kappa
      • CWAttackConfig.learning_rate
      • CWAttackConfig.max_iterations
      • CWAttackConfig.patience
  • advsecurenet.shared.types.configs.attack_configs.decision_boundary_attack_config module
    • DecisionBoundaryAttackConfig
      • DecisionBoundaryAttackConfig.early_stopping
      • DecisionBoundaryAttackConfig.early_stopping_patience
      • DecisionBoundaryAttackConfig.early_stopping_threshold
      • DecisionBoundaryAttackConfig.initial_delta
      • DecisionBoundaryAttackConfig.initial_epsilon
      • DecisionBoundaryAttackConfig.max_delta_trials
      • DecisionBoundaryAttackConfig.max_epsilon_trials
      • DecisionBoundaryAttackConfig.max_initialization_trials
      • DecisionBoundaryAttackConfig.max_iterations
      • DecisionBoundaryAttackConfig.step_adapt
      • DecisionBoundaryAttackConfig.targeted
      • DecisionBoundaryAttackConfig.verbose
  • advsecurenet.shared.types.configs.attack_configs.deepfool_attack_config module
    • DeepFoolAttackConfig
      • DeepFoolAttackConfig.max_iterations
      • DeepFoolAttackConfig.num_classes
      • DeepFoolAttackConfig.overshoot
  • advsecurenet.shared.types.configs.attack_configs.fgsm_attack_config module
    • FgsmAttackConfig
      • FgsmAttackConfig.epsilon
  • advsecurenet.shared.types.configs.attack_configs.lots_attack_config module
    • LotsAttackConfig
      • LotsAttackConfig.deep_feature_layer
      • LotsAttackConfig.mode
      • LotsAttackConfig.epsilon
      • LotsAttackConfig.learning_rate
      • LotsAttackConfig.max_iterations
      • LotsAttackConfig.verbose
      • LotsAttackConfig.device
      • LotsAttackConfig.deep_feature_layer
      • LotsAttackConfig.epsilon
      • LotsAttackConfig.learning_rate
      • LotsAttackConfig.max_iterations
      • LotsAttackConfig.mode
      • LotsAttackConfig.verbose
    • LotsAttackMode
      • LotsAttackMode.ITERATIVE
      • LotsAttackMode.SINGLE
  • advsecurenet.shared.types.configs.attack_configs.pgd_attack_config module
    • PgdAttackConfig
      • PgdAttackConfig.alpha
      • PgdAttackConfig.epsilon
      • PgdAttackConfig.num_iter
• advsecurenet.shared.types.configs.defense_configs package
  • advsecurenet.shared.types.configs.defense_configs.adversarial_training_config module
    • AdversarialTrainingConfig
      • AdversarialTrainingConfig.attacks
      • AdversarialTrainingConfig.models

---

advsecurenet.shared.types.configs.configs module

class advsecurenet.shared.types.configs.configs.ConfigType(value)

Bases: Enum

Enum for configuration types that can be loaded in the CLI.

ADVERSARIAL_EVALUATION = 'adversarial_evaluation'

ADVERSARIAL_TRAINING = 'adversarial_training'

ATTACK = 'attack'

DATASET = 'dataset'

DEFENSE = 'defense'

EVALUATION = 'evaluation'

MODEL = 'model'

TEST = 'test'

TRAIN = 'train'

advsecurenet.shared.types.configs.dataloader_config module

class advsecurenet.shared.types.configs.dataloader_config.DataLoaderConfig(dataset: BaseDataset, batch_size: int | None = 16, num_workers: int | None = 4, shuffle: bool | None = True, sampler: DistributedSampler | None = None, pin_memory: bool | None = True, drop_last: bool | None = False)

Bases: object

This dataclass is used to store the configuration of the data loader.

batch_size: int | None = 16

dataset: BaseDataset

drop_last: bool | None = False

num_workers: int | None = 4

pin_memory: bool | None = True

sampler: DistributedSampler | None = None

shuffle: bool | None = True

advsecurenet.shared.types.configs.device_config module

class advsecurenet.shared.types.configs.device_config.DeviceConfig(use_ddp: bool | None = False, processor: str | None = 'cpu', gpu_ids: str | None = None)

Bases: object

This dataclass is used to store the configuration of the device.

gpu_ids: str | None = None

processor: str | None = 'cpu'

use_ddp: bool | None = False

advsecurenet.shared.types.configs.model_config module

class advsecurenet.shared.types.configs.model_config.BaseModelConfig(*, model_name: str, num_classes: int | None = 1000, num_input_channels: int | None = 3, pretrained: bool | None = False)

Bases: object

Base configuration class for different model configurations.

model_name: str

num_classes: int | None = 1000

num_input_channels: int | None = 3

pretrained: bool | None = False

class advsecurenet.shared.types.configs.model_config.CreateModelConfig(model_arch_path: str | None = None, model_weights_path: str | None = None, custom_models_path: str | None = 'CustomModels', weights: str | None = 'IMAGENET1K_V1', is_external: bool = False, random_seed: int | None = None, *, model_name: str, num_classes: int | None = 1000, num_input_channels: int | None = 3, pretrained: bool | None = False)

Bases: StandardModelConfig, CustomModelConfig, ExternalModelConfig

Config parameters for creating a model in the model factory.

is_external: bool = False

random_seed: int | None = None

class advsecurenet.shared.types.configs.model_config.CustomModelConfig(custom_models_path: str | None = 'CustomModels', *, model_name: str, num_classes: int | None = 1000, num_input_channels: int | None = 3, pretrained: bool | None = False)

Bases: BaseModelConfig

Configuration for a custom model.

custom_models_path: str | None = 'CustomModels'

class advsecurenet.shared.types.configs.model_config.ExternalModelConfig(model_arch_path: str | None = None, model_weights_path: str | None = None, *, model_name: str, num_classes: int | None = 1000, num_input_channels: int | None = 3, pretrained: bool | None = False)

Bases: BaseModelConfig

Configuration for an external model.

model_arch_path: str | None = None

model_weights_path: str | None = None

class advsecurenet.shared.types.configs.model_config.StandardModelConfig(weights: str | None = 'IMAGENET1K_V1', *, model_name: str, num_classes: int | None = 1000, num_input_channels: int | None = 3, pretrained: bool | None = False)

Bases: BaseModelConfig

Configuration for a standard model.

weights: str | None = 'IMAGENET1K_V1'

advsecurenet.shared.types.configs.preprocess_config module

class advsecurenet.shared.types.configs.preprocess_config.PreprocessConfig(steps: list[PreprocessStep] | None = None)

Bases: object

This dataclass is used to store the configuration of the preprocessing pipeline.

steps: list[PreprocessStep] | None = None

class advsecurenet.shared.types.configs.preprocess_config.PreprocessStep(name: str, params: dict | None = None)

Bases: object

This dataclass is used to store the configuration of a preprocessing step.

name: str

params: dict | None = None

advsecurenet.shared.types.configs.test_config module

class advsecurenet.shared.types.configs.test_config.TestConfig(model: BaseModel, test_loader: DataLoader, criterion: str | Module = 'cross_entropy', processor: device | None = device(type='cpu'), topk: int = 1)

Bases: object

This dataclass is used to store the configuration of the test CLI.

criterion: str | Module = 'cross_entropy'

model: BaseModel

processor: device | None = device(type='cpu')

test_loader: DataLoader

topk: int = 1

advsecurenet.shared.types.configs.train_config module

class advsecurenet.shared.types.configs.train_config.CheckpointConfig(save_checkpoint: bool = False, save_checkpoint_path: str | None = None, save_checkpoint_name: str | None = None, checkpoint_interval: int = 1, load_checkpoint: bool = False, load_checkpoint_path: str | None = None)

Bases: object

Configuration class for the checkpoint.

checkpoint_interval: int = 1

load_checkpoint: bool = False

load_checkpoint_path: str | None = None

save_checkpoint: bool = False

save_checkpoint_name: str | None = None

save_checkpoint_path: str | None = None

class advsecurenet.shared.types.configs.train_config.FinalModelConfig(save_final_model: bool = False, save_model_path: str | None = None, save_model_name: str | None = None)

Bases: object

Configuration class for the final model.

save_final_model: bool = False

save_model_name: str | None = None

save_model_path: str | None = None

class advsecurenet.shared.types.configs.train_config.ModelConfig(model: Module | None = None)

Bases: object

Configuration class for the model.

model: Module = None

class advsecurenet.shared.types.configs.train_config.OptimizationConfig(optimizer: str | Optimizer = 'adam', optimizer_kwargs: dict | None = None, scheduler: str | Module | None = None, scheduler_kwargs: dict | None = None)

Bases: object

Configuration class for the optimization process.

optimizer: str | Optimizer = 'adam'

optimizer_kwargs: dict | None = None

scheduler: str | Module | None = None

scheduler_kwargs: dict | None = None

class advsecurenet.shared.types.configs.train_config.TrainConfig(use_ddp: bool | None = False, processor: str | None = 'cpu', gpu_ids: str | None = None, save_final_model: bool = False, save_model_path: str | None = None, save_model_name: str | None = None, save_checkpoint: bool = False, save_checkpoint_path: str | None = None, save_checkpoint_name: str | None = None, checkpoint_interval: int = 1, load_checkpoint: bool = False, load_checkpoint_path: str | None = None, optimizer: str | Optimizer = 'adam', optimizer_kwargs: dict | None = None, scheduler: str | Module | None = None, scheduler_kwargs: dict | None = None, train_loader: DataLoader | None = None, criterion: str | Module = 'cross_entropy', epochs: int = 10, learning_rate: float = 0.001, verbose: bool = False, model: Module | None = None)

Bases: ModelConfig, TrainingProcessConfig, OptimizationConfig, CheckpointConfig, FinalModelConfig, DeviceConfig

Dataclass to store the overall training configuration by aggregating other configurations.

class advsecurenet.shared.types.configs.train_config.TrainingProcessConfig(train_loader: DataLoader | None = None, criterion: str | Module = 'cross_entropy', epochs: int = 10, learning_rate: float = 0.001, verbose: bool = False)

Bases: object

Configuration class for the training process.

criterion: str | Module = 'cross_entropy'

epochs: int = 10

learning_rate: float = 0.001

train_loader: DataLoader = None

verbose: bool = False

---