README.md

---
language: en
tags:
- deberta
- fill-mask
license: mit
pipeline_tag: text-generation
---

# DeBERTa (1.5B) fixed version

This is [**deberta-v2-xxlarge**](https://huggingface.co/microsoft/deberta-v2-xxlarge) updated to implement the `AutoModelForCausalLM` class, enabling it to generate text. This implementation is based on our paper [**"BERTs are Generative In-Context Learners"**](https://arxiv.org/abs/2406.04823).

This repository also fixes three bugs in [the original HF implementation of DeBERTa](https://huggingface.co/microsoft/deberta-v2-xxlarge):
1. We fixed the incorrect name of the output embedding weights in the checkpoint file;
2. We fixed the implementation of the enhanced mask decoder (EMD), based on [the original GitHub repository](https://github.com/microsoft/DeBERTa);
3. We clamp the positional embeddings so that they work with long sequence lengths.

## Example code

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ltg/deberta-xxlarge-fixed", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("ltg/deberta-xxlarge-fixed", trust_remote_code=True).cuda().eval()

prompt = """German: Hallo, wie geht es Ihnen heute?
English:"""
prompt = prompt.replace('\n', '\\n ')
input_ids = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).input_ids.cuda()

prediction = model.generate(
    input_ids,
    num_beams=4,
    do_sample=False,
    use_cache=None,
    max_new_tokens=64,
    eos_token_id=tokenizer(".\\", add_special_tokens=False).input_ids[1:]
)
prediction = prediction[0, input_ids.size(1):]
prediction = tokenizer.decode(prediction).rstrip('\\')

# Expected output: "Hello, how are you doing today?"
print(prediction)
```


## Citation

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

```bibtex
@inproceedings{
  samuel2024berts,
  title={{BERT}s are Generative In-Context Learners},
  author={David Samuel},
  booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
  year={2024},
  url={https://openreview.net/forum?id=BCA9NMZkLS}
}
```

``` bibtex
@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}
}
```