---
license: apache-2.0
datasets:
- openbmb/Ultra-FineWeb
language:
- en
- zh
base_model:
- BlinkDL/rwkv7-g1
pipeline_tag: translation
tags:
- RWKV_V7
- Englisg->Chinese
- 0.4B
- 1.5B
---


## 🧠 Model Overview

- This project provides an **English-to-Chinese translation model** based on the **RWKV-V7 architecture**, with approximately **0.4 billion parameters** and **1.5 billion parameters**.
- Model weight fine-tuning base on [https://huggingface.co/BlinkDL/rwkv7-g1](https://huggingface.co/BlinkDL/rwkv7-g1)
- The model has been fully fine-tuned on translation tasks and demonstrates strong performance across various domains, especially in handling long sentences, technical terminology, and culturally nuanced expressions.
- Unlike traditional Transformer-based models, RWKV combines the sequential state-passing mechanism of RNNs with the parallel training capabilities of Transformers. This unique design enables efficient inference while maintaining powerful sequence modeling abilities, making it ideal for deployment on resource-constrained environments such as mobile devices, embedded systems, or edge computing platforms.

### 📦 Install Dependencies
#### 🟢 For Nvidia CUDA 
```bash
pip install torch rwkv gradio python-docx PyPDF2 chardet
```
#### 🔴 For AMD ROCm 
- set ```os.environ["RWKV_CUDA_ON"] = '0' ```
```bash
pip install torch --index-url https://download.pytorch.org/whl/rocm6.3
pip install rwkv python-docx PyPDF2 chardet
```

### 😜 Run The demo 
- Change line 20 in ```webui_new.py``` to you own model weights path
```bash
python webui_new.py 
```

## ⚠️ Notice

~~- This model currently supports **English → Chinese** translation only.~~
- Now it support **English → Chinese** & **Chinese → English** ~~~


## 💡 Key Advantages

- ✅ **Lightweight and Deployment-Friendly**: Achieves high-quality translation with only 0.4B / 1.5B parameters.
- ✅ **Strong Long-Context Modeling**: Supports input lengths up to 4096 tokens.
- ✅ **Low Memory Footprint**: Ideal for edge devices, mobile apps, and embedded systems.
- ✅ **Multilingual Potential**: Built upon a multilingual pre-training foundation, future versions may support more language pairs.

## 🎁 Recommended Resources

- 📘 [Official RWKV Repo](https://github.com/BlinkDL/RWKV-LM)
- 🧪 [Official RWKV Website](https://www.rwkv.cn/)
- 🧰 [Official RWKV project collection](https://github.com/RWKV-Vibe)
- 🐦 [Official fine-tuning Repo](https://github.com/JL-er/RWKV-PEFT)
- 🤖 [RWKV Runner](https://github.com/josStorer/RWKV-Runner)
- 👀 [AI00 Web Server](https://github.com/Ai00-X/ai00_server)

## 🧩 Developer Info

- **Developer**: Alic Li
- **GitHub**: [https://github.com/Alic-Li](https://github.com/Alic-Li)
- **Contact**: [alic2591709191@gmail.com](alic2591709191@gmail.com)