| --- |
| 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). |
