Towards Optimal Adapter Placement for Efficient Transfer Learning
read the original abstract
Parameter-efficient transfer learning (PETL) aims to adapt pre-trained models to new downstream tasks while minimizing the number of fine-tuned parameters. Adapters, a popular approach in PETL, inject additional capacity into existing networks by incorporating low-rank projections, achieving performance comparable to full fine-tuning with significantly fewer parameters. This paper investigates the relationship between the placement of an adapter and its performance. We observe that adapter location within a network significantly impacts its effectiveness, and that the optimal placement is task-dependent. To exploit this observation, we introduce an extended search space of adapter connections, including long-range and recurrent adapters. We demonstrate that even randomly selected adapter placements from this expanded space yield improved results, and that high-performing placements often correlate with high gradient rank. Our findings reveal that a small number of strategically placed adapters can match or exceed the performance of the common baseline of adding adapters in every block, opening a new avenue for research into optimal adapter placement strategies.
This paper has not been read by Pith yet.
Forward citations
Cited by 3 Pith papers
-
GD-FPS: Growth-Driven Feedforward Parameter Selection for Efficient Fine-Tuning
GD-FPS is a gradient-free, forward-pass-only parameter selection method for PEFT that identifies important weights by scaling magnitudes with relative activation growth against a pre-training anchor, matching or beati...
-
The Topological Trouble With Transformers
Transformers face a topological limitation in dynamic state tracking because their feedforward architecture pushes evolving state representations deeper into layers until depth is exhausted, requiring a shift to recur...
-
The Topological Trouble With Transformers
An intermediate-fusion model concatenating ResNet-18 image embeddings with MLP-derived clinical embeddings reports improved mitosis-class AUC, but the image and tabular datasets are unpaired, undermining the fusion claim.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.