#lora #adaptation #continuous-learning #online-learning

haagenti-learning

Continuous learning and online adaptation for models

1 unstable release

new 0.1.0 Jan 27, 2026

#1568 in Algorithms


Used in haagenti

MIT/Apache

110KB
2.5K SLoC

Continuous learning and online adaptation

This crate provides continuous learning capabilities:

  • LoRA adapter training and merging
  • Experience replay buffer for online learning
  • Elastic Weight Consolidation (EWC) for catastrophic forgetting prevention
  • Progressive layer unfreezing
  • Learning rate scheduling
  • Gradient accumulation and checkpointing

Dependencies

~11–25MB
~230K SLoC