File size: 2,171 Bytes
5674138
 
 
 
 
 
d9f70c4
 
5674138
d9f70c4
 
 
 
 
 
 
 
53925ef
d9f70c4
 
53925ef
d9f70c4
 
 
 
 
 
 
 
 
 
 
bad5285
 
 
 
d9f70c4
 
 
 
bad5285
d9f70c4
 
 
 
 
 
 
 
 
 
 
89a5602
 
 
 
 
d9f70c4
 
 
 
 
 
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
---
license: apache-2.0
language:
- zh
base_model:
- hfl/chinese-lert-base
tags:
- punctuation-restoration
---
<div align="center">
<h1>FireRedChat-punc</h1>
</div>
<div align="center">
  <a href="https://fireredteam.github.io/demos/firered_chat/">Demo</a><a href="https://arxiv.org/pdf/2509.06502">FireRedChat Paper</a><a href="https://huggingface.co/FireRedTeam">Huggingface</a>  
</div>

## Descriptions
FireRedChat-punc is a fine-tuned `hfl/chinese-lert-base` model designed for punctuation restoration, primarily for post-processing in [FireRedASR](https://github.com/FireRedTeam/FireRedASR). 

The model restores the following punctuation marks: [, 。 ? !]. It supports both Chinese and English text, enhancing the readability of transcribed text.

## Roadmap
- [x] 2025/09
  - [x] Release the fine-tuned punctuation restoration model.

## Usage

RedPost source code [github](https://github.com/FireRedTeam/FireRedChat/tree/main/fireredasr-server/server/redpost)
Below is an example of how to use the FireRedChat-punc model for punctuation restoration:

```bash
git clone https://huggingface.co/hfl/chinese-lert-base FireRedChat-punc/chinese-lert-base
```

```python
import os
from redpost import RedPost, RedPostConfig

punc_model_dir = "./FireRedChat-punc"
post_config = RedPostConfig(
    use_gpu=True,
    sentence_max_length=30
)
post_model = RedPost.from_pretrained(punc_model_dir, post_config)
batch_post_results = post_model.process([text], ["text"])
text = "".join([r["punc_text"] for r in batch_post_results])
text = re.sub("<unk>|<UNK>|\[unk\]|\[UNK\]", "", text)
print(text)
```

## Use with FireRedASR

This punctuation restoration model can be used together with FireRedASR, refer to [fireredasr-server](https://github.com/FireRedTeam/FireRedChat/blob/main/fireredasr-server/README.md) for setup instructions.
https://github.com/FireRedTeam/FireRedChat/tree/main/fireredasr-server

## License
The model and source code are licensed under the Apache-2.0 license.

### Acknowledgment
- Base model: `hfl/chinese-lert-base` (license: apache-2.0)
- Designed for integration with [FireRedASR](https://github.com/FireRedTeam/FireRedASR).