README.md

---
language: en
tags: 
  - deberta
  - deberta-v3
thumbnail: https://huggingface.co/front/thumbnails/microsoft.png
license: mit
---

## DeBERTa: Decoding-enhanced BERT with Disentangled Attention

[DeBERTa](https://arxiv.org/abs/2006.03654) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. With those two improvements, DeBERTa out perform RoBERTa on a majority of NLU tasks with 80GB training data. 

Please check the [official repository](https://github.com/microsoft/DeBERTa) for more details and updates.

In DeBERTa V3 we replaced MLM objective with RTD(Replaced Token Detection) objective which was first introduced by ELECTRA for pre-training. The new objective significantly improves the model performance. Please check appendix A11 in our paper [DeBERTa](https://arxiv.org/abs/2006.03654) for more details.

This is the DeBERTa V3 small model with 6 layers, 768 hidden size. Total parameters is 143M while Embedding layer take about 98M due to the usage of 128k vocabulary. It's trained with 160GB data.

#### Fine-tuning on NLU tasks

We present the dev results on SQuAD 1.1/2.0 and MNLI tasks.

| Model             | SQuAD 1.1 | SQuAD 2.0 | MNLI-m |
|-------------------|-----------|-----------|--------|
| RoBERTa-base      | 91.5/84.6 | 83.7/80.5 | 87.6   |
| XLNet-base        | -/-       | -/80.2    | 86.8   |
|DeBERTa-base	|93.1/87.2|	86.2/83.1|	88.8|
| **DeBERTa-v3-small**  | -/- | -/- | 88.1   |
| DeBERTa-v3-small+SiFT  | -/- | -/- | 88.8   |


#### Fine-tuning with HF transformers

```bash
#!/bin/bash

cd transformers/examples/pytorch/text-classification/

pip install datasets
export TASK_NAME=mnli

output_dir="ds_results"

num_gpus=8

batch_size=8

python -m torch.distributed.launch --nproc_per_node=${num_gpus} \
  run_glue.py \
  --model_name_or_path microsoft/deberta-v3-small \
  --task_name $TASK_NAME \
  --do_train \
  --do_eval \
  --evaluation_strategy steps \
  --max_seq_length 256 \
  --warmup_steps 1000 \
  --per_device_train_batch_size ${batch_size} \
  --learning_rate 3e-5 \
  --num_train_epochs 3 \
  --output_dir $output_dir \
  --overwrite_output_dir \
  --logging_steps 1000 \
  --logging_dir $output_dir

```

### Citation

If you find DeBERTa useful for your work, please cite the following paper:

``` latex
@inproceedings{
he2021deberta,
title={DEBERTA: DECODING-ENHANCED BERT WITH DISENTANGLED ATTENTION},
author={Pengcheng He and Xiaodong Liu and Jianfeng Gao and Weizhu Chen},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=XPZIaotutsD}
}
```