File size: 1,786 Bytes
0ecc44a
 
 
bc65833
843d4d9
97311a6
 
 
 
 
 
 
 
 
 
 
a0cdf41
97311a6
 
 
 
 
 
cff2856
 
 
4445e55
97311a6
 
 
1909856
 
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
---
license: apache-2.0
---

# Korean Character BERT Model (small)

Welcome to the repository of the Korean Character (syllable-level) BERT Model, a compact and efficient transformer-based model designed specifically for Korean language processing tasks. This model takes a unique approach by tokenizing text at the syllable level, catering to the linguistic characteristics of the Korean language.

## Features

- Vocabulary Size: The model utilizes a vocabulary of 7,477 tokens, focusing on Korean syllables. This streamlined vocabulary size allows for efficient processing while maintaining the ability to capture the nuances of the Korean language.
- Transformer Encoder Layers: It employs a simplified architecture with only 3 transformer encoder layers. This design choice strikes a balance between model complexity and computational efficiency, making it suitable for a wide range of applications, from mobile devices to server environments.
- License: This model is open-sourced under the Apache License 2.0, allowing for both academic and commercial use while ensuring that contributions and improvements are shared within the community.

## Getting Started

```python
# Load model directly
from transformers import AutoTokenizer, AutoModelForMaskedLM

tokenizer = AutoTokenizer.from_pretrained("MrBananaHuman/char_ko_bert_small")
model = AutoModelForMaskedLM.from_pretrained("MrBananaHuman/char_ko_bert_small")
```

## Fine-tuning example

- [Named entity recognition](https://colab.research.google.com/drive/1WirfVhJIbKH70stuLRPhiPr2CexZiGuP?usp=sharing)

## Contact

For any questions or inquiries, please reach out to me at mrbananahuman.kim@gmail.com    
I'm always happy to discuss the model, potential collaborations, or any other inquiries related to this project.