Adapters provides a unified interface for efficient fine-tuning and modular transfer learning, supporting a myriad of features like full-precision or quantized training (e.g. Q-LoRA, Q-Bottleneck ...