avalanche.benchmarks.scenarios.OnlineCLExperience
- class avalanche.benchmarks.scenarios.OnlineCLExperience(*, dataset: TCLDataset, origin_experience: DatasetExperience, is_first_subexp: bool = False, is_last_subexp: bool = False, sub_stream_length: int | None = None, access_task_boundaries: bool = False)[source]
Online CL (OCL) Experience.
OCL experiences are created by splitting a larger experience. Therefore, they keep track of the original experience for logging purposes.
- __init__(*, dataset: TCLDataset, origin_experience: DatasetExperience, is_first_subexp: bool = False, is_last_subexp: bool = False, sub_stream_length: int | None = None, access_task_boundaries: bool = False)[source]
A class representing a continual learning experience in an online setting.
- Parameters:
current_experience (int) – The index of the current experience.
dataset (TCLDataset) – The dataset containing the experience.
origin_experience (DatasetExperience) – The original experience from which this experience was derived.
is_first_subexp (bool, optional) – Whether this is the first sub-experience.
is_last_subexp (bool, optional) – Whether this is the last sub-experience.
sub_stream_length (int, optional) – The length of the sub-stream.
access_task_boundaries (bool, optional) – Whether to access task boundaries.
Methods
__init__
(*, dataset, origin_experience[, ...])A class representing a continual learning experience in an online setting.
eval
()Return inference experience.
logging
()Return logging experience.
no_attribute_masking
()train
()Return training experience.
Attributes
are_attributes_masked
current_experience
dataset
origin_stream