metadata

tags:
  - adapter-transformers
  - xlm-roberta
datasets:
  - UKPLab/m2qa

M2QA Adapter: Language Adapter for MAD-X² Setup

This adapter is part of the M2QA publication to achieve language and domain transfer via adapters.
📃 Paper: https://arxiv.org/abs/2407.01091
🏗️ GitHub repo: https://github.com/UKPLab/m2qa
💾 Hugging Face Dataset: https://huggingface.co/UKPLab/m2qa

Important: This adapter only works together with the MAD-X-2 domain and QA head adapter.

This adapter for the xlm-roberta-base model that was trained using the Adapters library. For detailed training details see our paper or GitHub repository: https://github.com/UKPLab/m2qa. You can find the evaluation results for this adapter on the M2QA dataset in the GitHub repo and in the paper.

Usage

First, install adapters:

pip install -U adapters

Now, the adapter can be loaded and activated like this:

from adapters import AutoAdapterModel
from adapters.composition import Stack

model = AutoAdapterModel.from_pretrained("xlm-roberta-base")

# 1. Load language adapter
language_adapter_name = model.load_adapter("AdapterHub/m2qa-xlm-roberta-base-mad-x-2-turkish") 

# 2. Load domain adapter
domain_adapter_name = model.load_adapter("AdapterHub/m2qa-xlm-roberta-base-mad-x-2-creative-writing")

# 3. Load QA head adapter
qa_adapter_name = model.load_adapter("AdapterHub/m2qa-xlm-roberta-base-mad-x-2-qa-head")

# 4. Activate them via the adapter stack
model.active_adapters = Stack(language_adapter_name, domain_adapter_name, qa_adapter_name)

See our repository for more information: See https://github.com/UKPLab/m2qa/tree/main/Experiments/mad-x-2

Contact

Leon Engländer:

Citation

@article{englaender-etal-2024-m2qa,
    title="M2QA: Multi-domain Multilingual Question Answering",
    author={Engl{\"a}nder, Leon and
        Sterz, Hannah and
        Poth, Clifton and
        Pfeiffer, Jonas and
        Kuznetsov, Ilia and
        Gurevych, Iryna},
    journal={arXiv preprint},
    url="https://arxiv.org/abs/2407.01091",
    month = jul,
    year="2024"
}