Rapid Switching and Multi-Adapter Fusion via Sparse High Rank Adapters
This work addresses efficient adaptation for large models, offering incremental improvements in parameter efficiency and fusion capabilities.
The paper tackles the problem of efficient fine-tuning for large models by proposing Sparse High Rank Adapters (SHiRA), which directly fine-tune 1-2% of base model weights, resulting in no inference overhead, rapid switching, and reduced concept-loss in multi-adapter fusion, outperforming Low Rank Adaptation (LoRA) in experiments on LVMs and LLMs.
In this paper, we propose Sparse High Rank Adapters (SHiRA) that directly finetune 1-2% of the base model weights while leaving others unchanged, thus, resulting in a highly sparse adapter. This high sparsity incurs no inference overhead, enables rapid switching directly in the fused mode, and significantly reduces concept-loss during multi-adapter fusion. Our extensive experiments on LVMs and LLMs demonstrate that finetuning merely 1-2% parameters in the base model is sufficient for many adapter tasks and significantly outperforms Low Rank Adaptation (LoRA). We also show that SHiRA is orthogonal to advanced LoRA methods such as DoRA and can be easily combined with existing techniques.