avalanche.benchmarks.scenarios.ExModelCLScenario

class avalanche.benchmarks.scenarios.ExModelCLScenario(original_benchmark: CLScenario, expert_models: List[Module])[source]

Ex-Model CL Scenario.

Ex-Model Continual Learning (ExML) is a continual learning scenario where the CL agent learns from a stream of pretrained models instead of raw data. These approach allows to integrate knowledge from different CL agents or pretrained models.

Reference: Carta, A., Cossu, A., Lomonaco, V., & Bacciu, D. (2021). Ex-Model: Continual Learning from a Stream of Trained Models. arXiv preprint arXiv:2112.06511. https://arxiv.org/abs/2112.06511

__init__(original_benchmark: CLScenario, expert_models: List[Module])[source]

Init.

Parameters
  • original_benchmark – a reference to the original benchmark containing the stream of experiences used to train the experts.

  • expert_models – pretrained models. The model in position i must be trained on the i-th experience of the train stream of original_benchmark.

Methods

__init__(original_benchmark, expert_models)

Init.

Attributes

streams