In this update, we support the state-of-the-art methods of 2023:
MEMO: A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning. ICLR 2023 Spotlight [paper]BEEF: BEEF: Bi-Compatible Class-Incremental Learning via Energy-Based Expansion and Fusion. ICLR 2023 [paper]SimpleCIL: Revisiting Class-Incremental Learning with Pre-Trained Models: Generalizability and Adaptivity are All You Need. arXiv 2023 [paper]
Cheers!