Edit on GitHub

model = AutoAdapterModel.from_pretrained("bert-base-multilingual-cased")
config = AdapterConfig.load("houlsby", non_linearity="gelu", reduction_factor=2)
model.load_adapter("ar/wiki@ukp", config=config)

Description

Houlsby Adapter trained with Masked Language Modelling on Arabic Wikipedia Articles for 250k steps and a batch size of 64.

Properties

Pre-trained model
bert-base-multilingual-cased
Adapter type
Prediction Head
  No
Task
Arabic
Dataset

Architecture

Name
houlsby
Non-linearity
gelu
Reduction factor
2
{
  "ln_after": false,
  "ln_before": false,
  "mh_adapter": true,
  "output_adapter": true,
  "adapter_residual_before_ln": false,
  "non_linearity": "gelu",
  "original_ln_after": true,
  "original_ln_before": false,
  "reduction_factor": 2,
  "residual_before_ln": true
}

Author

  Name
Jonas Pfeiffer
  GitHub
  Twitter

Versions

Identifier Comment Score Download
nd DEFAULT
wd

Citations

Adapter
@article{pfeiffer20madx,
  title={{MAD-X}: An {A}dapter-based {F}ramework for {M}ulti-task {C}ross-lingual {T}ransfer},
  author={Pfeiffer, Jonas and Vuli\'{c}, Ivan and Gurevych, Iryna and Ruder, Sebastian},
  journal={arXiv preprint},
  year={2020},
  url={https://arxiv.org/pdf/2005.00052.pdf},
}
Architecture
@misc{houlsby2019parameterefficient,
  title={Parameter-Efficient Transfer Learning for NLP},
  author={Neil Houlsby and Andrei Giurgiu and Stanislaw Jastrzebski and Bruna Morrone and Quentin de Laroussilhe and Andrea Gesmundo and Mona Attariyan and Sylvain Gelly},
  year={2019},
  eprint={1902.00751},
  archivePrefix={arXiv},
  primaryClass={cs.LG}
}