Edit on GitHub

model = AutoAdapterModel.from_pretrained("aubmindlab/bert-base-arabert")
config = AdapterConfig.load("pfeiffer")
model.load_adapter("dialect/arabic@mapmeld", config=config)

Description

Adapter for AraBERT (aubmindlab/bert-base-arabert) trained to classify Arabic by dialect {0=Egyptian, 1=Gulf, 2=Levantine, 3=Maghrebi, 4=MSA} Trained for 3 epochs on 85k samples (+ 28k test set) from University of British Columbia and John Hopkins University.

Properties

Pre-trained model
aubmindlab/bert-base-arabert
Adapter type
Prediction Head
  Yes
Task
Dialect Detection

Architecture

Name
pfeiffer
Non-linearity
relu
Reduction factor
16
{
  "ln_after": false,
  "ln_before": false,
  "mh_adapter": false,
  "output_adapter": true,
  "adapter_residual_before_ln": false,
  "non_linearity": null,
  "original_ln_after": true,
  "original_ln_before": true,
  "reduction_factor": null,
  "residual_before_ln": true
}

Author

  Name
Nick Doiron
  GitHub
  Twitter

Versions

Identifier Comment Score Download
1
2 DEFAULT

Citations

Architecture
@misc{pfeiffer2020adapterfusion,
  title={AdapterFusion: Non-Destructive Task Composition for Transfer Learning},
  author={Jonas Pfeiffer and Aishwarya Kamath and Andreas Rücklé and Kyunghyun Cho and Iryna Gurevych},
  year={2020},
  eprint={2005.00247},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}