Deberta V3 Small

🚀 DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing

DeBERTaV3 enhances the efficiency of DeBERTa through ELECTRA-Style pre-training with Gradient Disentangled Embedding Sharing, significantly improving model performance on downstream tasks.

🚀 Quick Start

DeBERTa enhances the BERT and RoBERTa models by utilizing disentangled attention and an enhanced mask decoder. With these two improvements, DeBERTa outperforms RoBERTa on a majority of NLU tasks using 80GB of training data.

In DeBERTa V3, we further improved the efficiency of DeBERTa using ELECTRA-Style pre-training with Gradient Disentangled Embedding Sharing. Compared to DeBERTa, our V3 version significantly improves the model performance on downstream tasks. You can find more technical details about the new model in our paper.

Please check the official repository for more implementation details and updates.

The DeBERTa V3 small model has 6 layers and a hidden size of 768. It has 44M backbone parameters with a vocabulary containing 128K tokens, which introduces 98M parameters in the Embedding layer. This model was trained using 160GB of data, the same as DeBERTa V2.

✨ Features

Fine-tuning on NLU tasks

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

💻 Usage Examples

Fine-tuning with HF transformers

#!/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 1500 \
  --per_device_train_batch_size ${batch_size} \
  --learning_rate 4.5e-5 \
  --num_train_epochs 3 \
  --output_dir $output_dir \
  --overwrite_output_dir \
  --logging_steps 1000 \
  --logging_dir $output_dir

📚 Documentation

Citation

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

@misc{he2021debertav3,
      title={DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing}, 
      author={Pengcheng He and Jianfeng Gao and Weizhu Chen},
      year={2021},
      eprint={2111.09543},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@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}
}

📄 License

This project is licensed under the MIT license.