---
language: 
  - zh

license: apache-2.0

tags:
  - bert

inference: true

widget:
- text: "生活的真谛是[MASK]。"
---
# Erlangshen-Deberta-XLarge-710M-Chinese，one model of [Fengshenbang-LM](https://github.com/IDEA-CCNL/Fengshenbang-LM)

The 710 million parameter deberta-V2 base model, using 180G Chinese data, 24 A100(40G) training for 21 days，which is a encoder-only transformer structure. Consumed totally 700M samples. Still training...

## Task Description

Erlangshen-Deberta-XLarge-710M-Chinese is pre-trained by bert like mask task from Deberta [paper](https://readpaper.com/paper/3033187248)

## Usage

```python
from transformers import AutoModelForMaskedLM, AutoTokenizer, FillMaskPipeline
import torch

tokenizer=AutoTokenizer.from_pretrained('IDEA-CCNL/Erlangshen-Deberta-XLarge-710M-Chinese', use_fast=false)
model=AutoModelForMaskedLM.from_pretrained('IDEA-CCNL/Erlangshen-Deberta-XLarge-710M-Chinese')
text = '生活的真谛是[MASK]。'
fillmask_pipe = FillMaskPipeline(model, tokenizer, device=-1)
print(fillmask_pipe(text, top_k=10))
```

## Finetune

We present the dev results on some tasks.

| Model                              | AFQMC|TNEWS1.1|IFLYTEK|OCNLI | CMNLI  |
| ---------------------------------- | ----- | ------ | ------ | ------ | ------ |
| RoBERTa-Large                       | 0.7488|0.5879|0.6152|0.777 | 0.814 |
| **Erlangshen-Deberta-XLarge-710M-Chinese** | 0.7549|0.5873|0.6177|0.8012|0.8389|

## Citation

If you find the resource is useful, please cite the following website in your paper.

```html
@misc{Fengshenbang-LM,
  title={Fengshenbang-LM},
  author={IDEA-CCNL},
  year={2022},
  howpublished={\url{https://github.com/IDEA-CCNL/Fengshenbang-LM}},
}
```