avalanche.benchmarks.scenarios.CLScenario
- class avalanche.benchmarks.scenarios.CLScenario(streams: List[CLStream])[source]
Continual Learning benchmark.
A Continual Learning benchmark is a container for a set of streams of experiences. It may also contain other additional data useful for evaluation purposes or logging.
The content of each experience depends on the underlying problem (data in a supervised problem, environments in reinforcement learning, and so on).
NOTE: benchmarks should not be used by training strategies since they provide access to past, current, and future data.
- __init__(streams: List[CLStream])[source]
Creates an instance of a Continual Learning benchmark.
- Parameters
streams – a list of streams.
Methods
__init__
(streams)Creates an instance of a Continual Learning benchmark.
Attributes
streams