avalanche.benchmarks.scenarios.CLStream

class avalanche.benchmarks.scenarios.CLStream(name: str, exps_iter: Iterable[TCLExperience], benchmark: CLScenario[TCLStream] | None = None, set_stream_info: bool = True)

A CL stream is a named iterator of experiences.

In general, many streams may be generator and not explicit lists to avoid keeping many objects in memory.

NOTE: streams should not be used by training strategies since they provide access to past, current, and future data.

__init__(name: str, exps_iter: Iterable[TCLExperience], benchmark: CLScenario[TCLStream] | None = None, set_stream_info: bool = True)

Creates an instance of a experience stream.

Parameters:

• name – The name of the stream.

• exps_iter – The iterable from which experiences will be obtained.

• benchmark – The benchmarks defining this stream.

• set_stream_info – If True, will set the current_experience and origin_stream fields on experience objects before returning them. Defaults to True.

Methods

__init__(name, exps_iter[, benchmark, ...])

Creates an instance of a experience stream.

Attributes

name	The name of the stream (for instance: "train", "test", "valid", ...).
exps_iter	The iterable from which experiences will be obtained.
benchmark	A reference to the benchmark.
set_stream_info	If True, will set the current_experience and origin_stream fields on experience objects before returning them.