File size: 3,585 Bytes
52978a1
 
e16fd71
 
 
a36c739
52978a1
 
 
 
2e588b2
52978a1
 
 
2e588b2
52978a1
2e588b2
66be1e4
2e588b2
bfb08ac
52978a1
 
 
2e588b2
1bd59c9
2e588b2
 
 
 
 
 
d92f88e
2e588b2
6f8dde6
2e588b2
52978a1
 
96d6beb
 
 
 
e2fd614
35addd0
e2fd614
96d6beb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2e588b2
96d6beb
2e588b2
81ffed8
96d6beb
 
 
 
 
 
 
52978a1
 
2e588b2
52978a1
7aad3ba
 
 
 
 
 
 
 
 
 
 
52978a1
 
 
 
 
 
 
 
 
2e588b2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
---
language: en
tags: 
  - deberta
  - deberta-v3
  - fill-mask
thumbnail: https://huggingface.co/front/thumbnails/microsoft.png
license: mit
---

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

[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. 

In [DeBERTa V3](https://arxiv.org/abs/2111.09543), 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 technique details about the new model from our [paper](https://arxiv.org/abs/2111.09543).

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

The DeBERTa V3 small model comes with 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 the 160GB data as DeBERTa V2.


#### Fine-tuning on NLU tasks

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

| Model             |Vocabulary(K)|Backbone #Params(M)| SQuAD 2.0(F1/EM) | MNLI-m/mm(ACC)|
|-------------------|----------|-------------------|-----------|----------|
| RoBERTa-base      |50     |86                 | 83.7/80.5 | 87.6/-   |
| XLNet-base        |32     |92                 | -/80.2    | 86.8/-   |
| ELECTRA-base      |30    |86                  | -/80.5    | 88.8/    |
| DeBERTa-base      |50     |100                |  86.2/83.1| 88.8/88.5|
| DeBERTa-v3-large|128|304                      |  91.5/89.0  | 91.8/91.9        |
| DeBERTa-v3-base |128|86                       | 88.4/85.4 | 90.6/90.7|
| **DeBERTa-v3-small**  |128|**44**                     | **82.8/80.4** | **88.3/87.7**|
| DeBERTa-v3-small+SiFT|128|22                 | -/-       | 88.8/88.5|


#### 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 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

```

### Citation

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

``` latex
@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}
}
```

``` 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}
}
```