File size: 1,113 Bytes
5c440c3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 |
## DynaBERT: Dynamic BERT with Adaptive Width and Depth
* DynaBERT can flexibly adjust the size and latency by selecting adaptive width and depth, and
the subnetworks of it have competitive performances as other similar-sized compressed models.
The training process of DynaBERT includes first training a width-adaptive BERT and then
allowing both adaptive width and depth using knowledge distillation.
* This code is modified based on the repository developed by Hugging Face: [Transformers v2.1.1](https://github.com/huggingface/transformers/tree/v2.1.1), and is released in [GitHub](https://github.com/huawei-noah/Pretrained-Language-Model/tree/master/DynaBERT).
### Reference
Lu Hou, Zhiqi Huang, Lifeng Shang, Xin Jiang, Xiao Chen, Qun Liu.
[DynaBERT: Dynamic BERT with Adaptive Width and Depth](https://arxiv.org/abs/2004.04037).
```
@inproceedings{hou2020dynabert,
title = {DynaBERT: Dynamic BERT with Adaptive Width and Depth},
author = {Lu Hou, Zhiqi Huang, Lifeng Shang, Xin Jiang, Xiao Chen, Qun Liu},
booktitle = {Advances in Neural Information Processing Systems},
year = {2020}
}
```
|