Multi-Head Adapter Routing for Cross-Task Generalization

• URL: http://arxiv.org/abs/2211.03831v3
• Date: Mon, 13 Nov 2023 15:09:59 GMT
• Title: Multi-Head Adapter Routing for Cross-Task Generalization
• Authors: Lucas Caccia, Edoardo Ponti, Zhan Su, Matheus Pereira, Nicolas Le Roux, Alessandro Sordoni
• Abstract summary: Polytropon learns an inventory of adapters and a routing function that selects a subset of adapters for each task during both pre-training and few-shot adaptation. We find that routing is most beneficial during multi-task pre-training rather than during few-shot adaptation.
• Score: 56.75667096355806
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Parameter-efficient fine-tuning (PEFT) for cross-task generalization consists in pre-training adapters on a multi-task training set before few-shot adaptation to test tasks. Polytropon [Ponti et al., 2023] ($\texttt{Poly}$) jointly learns an inventory of adapters and a routing function that selects a (variable-size) subset of adapters for each task during both pre-training and few-shot adaptation. In this paper, we investigate the role that adapter routing plays in its success and design new variants based on our findings. First, we build on the intuition that finer-grained routing provides more expressivity. Hence, we propose $\texttt{MHR}$ (Multi-Head Routing) which combines subsets of adapter parameters and outperforms $\texttt{Poly}$ under a comparable parameter budget; by only fine-tuning the routing function and not the adapters ($\texttt{MHR}$-$z$) we achieve competitive performance with extreme parameter efficiency. Second, we find that $\texttt{Poly}$/$\texttt{MHR}$ performance is a result of better multi-task optimization, rather than modular inductive biases that facilitate adapter recombination and local adaptation, as previously hypothesized. In fact, we find that $\texttt{MHR}$ exhibits high gradient alignment between training tasks. We find that routing is most beneficial during multi-task pre-training rather than during few-shot adaptation and propose $\texttt{MHR}$-$\mu$, which discards routing and fine-tunes the average of the pre-trained adapters on each downstream tasks. This establishes $\texttt{MHR}$-$\mu$ as an effective method for single-adapter fine-tuning. We also show that $\texttt{MHR}$-$\mu$ can be used as an effective zero-shot transfer method by training the average of the pre-trained adapters for a few additional steps on the multi-task training set: this yields gains up to 3% on absolute accuracy w.r.t. the baselines.

Related papers

• Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass [109.34230156454574]
  Large language models (LMs) are typically adapted to improve performance on new contexts. fine-tuning incurs significant training cost and prompting increases inference overhead. We introduce $GenerativeAdapter$, an effective and efficient adaptation method that directly maps new contexts to low-rank LM adapters.
  arXiv  Detail & Related papers  (2024-11-08T00:42:47Z)
• Hierarchical Recurrent Adapters for Efficient Multi-Task Adaptation of Large Speech Models [12.230087530720652]
  We introduce an adapter module that has a better efficiency in large scale multi-task adaptation scenario. The adapter consists of a single shared controller network and multiple task-level adapter heads.
  arXiv  Detail & Related papers  (2024-03-25T17:21:56Z)
• Prototype-based HyperAdapter for Sample-Efficient Multi-task Tuning [30.251155072822055]
  Prototype-based HyperAdapter (PHA) is a novel framework built on the adapter-tuning and hypernetwork. It introduces an instance-dense retriever and prototypical hypernetwork to generate conditional modules in a sample-efficient manner. We show that PHA strikes a better trade-off between trainable parameters, accuracy on stream tasks, and sample efficiency.
  arXiv  Detail & Related papers  (2023-10-18T02:42:17Z)
• MerA: Merging Pretrained Adapters For Few-Shot Learning [71.44422347502409]
  We propose textbftextttMerging Pretrained Adapters (MerA) that efficiently incorporates pretrained adapters to a single model through model fusion. Experiments on two PLMs demonstrate that MerA substantial improvements compared to both single adapters and AdapterFusion.
  arXiv  Detail & Related papers  (2023-08-30T12:10:17Z)
• Vision Transformer Adapters for Generalizable Multitask Learning [61.79647180647685]
  We introduce the first multitasking vision transformer adapters that learn generalizable task affinities. Our adapters can simultaneously solve multiple dense vision tasks in a parameter-efficient manner. In contrast to concurrent methods, we do not require retraining or fine-tuning whenever a new task or domain is added.
  arXiv  Detail & Related papers  (2023-08-23T18:40:48Z)
• Cross-Modal Adapter for Text-Video Retrieval [91.9575196703281]
  We present a novel $textbfCross-Modal Adapter$ for parameter-efficient fine-tuning. Inspired by adapter-based methods, we adjust the pre-trained model with a few parameterization layers. It achieves superior or comparable performance compared to fully fine-tuned methods on MSR-VTT, MSVD, VATEX, ActivityNet, and DiDeMo datasets.
  arXiv  Detail & Related papers  (2022-11-17T16:15:30Z)
• AdaMix: Mixture-of-Adapter for Parameter-efficient Tuning of Large Language Models [119.7093605087114]
  Fine-tuning large-scale pre-trained language models to downstream tasks require updating hundreds of millions of parameters. This not only increases the serving cost to store a large copy of the model weights for every task, but also exhibits instability during few-shot task adaptation. We introduce a new mechanism to improve adapter capacity without increasing parameters or computational cost by two key techniques.
  arXiv  Detail & Related papers  (2022-05-24T23:41:22Z)
• Parameter-efficient Multi-task Fine-tuning for Transformers via Shared Hypernetworks [37.2958914602899]
  We show that we can learn adapter parameters for all layers and tasks by generating them using shared hypernetworks. Experiments on the well-known GLUE benchmark show improved performance in multi-task learning while adding only 0.29% parameters per task.
  arXiv  Detail & Related papers  (2021-06-08T16:16:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.