---
library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen3-8B
language:
  - af  # Afrikaans
  - am  # Amharic
  - ar  # Arabic
  - en  # English
  - fr  # French
  - ha  # Hausa
  - ig  # Igbo
  - mg  # Malagasy (Plateau)
  - ny  # Nyanja
  - om  # Oromo
  - pt  # Portuguese
  - rw  # Kinyarwanda
  - sn  # Shona
  - so  # Somali
  - st  # Southern Sotho
  - sw  # Swahili
  - ti  # Tigrinya
  - tn  # Tswana
  - xh  # Xhosa
  - yo  # Yoruba
  - zu  # Zulu
pipeline_tag: text-generation
tags:
  - african-languages
  - multilingual
  - continued-pretraining
  - afrique-llm
  - qwen
---

# AfriqueQwen-8B

<p align="center">
  <img src="language_group_scores.pdf" alt="AfriqueLLM Evaluation Results" width="600"/>
</p>

## Model Overview

**AfriqueQwen-8B** is part of the **AfriqueLLM** suite—a collection of open language models adapted to **20 African languages** through continued pre-training (CPT) on **26B tokens**. This model is based on [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) and has been specifically adapted for improved performance on African languages while maintaining strong capabilities in high-resource languages.

Our experiments show that **Qwen 3 models achieve the best performance** among all base models tested, better preserving performance in high-resource languages after CPT and achieving strong results on long-context tasks such as document-level translation.

### Key Features

- **Type**: Causal Language Model (Base/Pre-trained)
- **Base Model**: Qwen 3 8B
- **Parameters**: 8B
- **Context Length**: 32,768 tokens (native)
- **Training Tokens**: 26B tokens of carefully curated multilingual data

## Supported Languages

AfriqueQwen-8B has been adapted for the following 20 African languages plus 4 high-resource languages:

| Language | Code | Family | Script |
|----------|------|--------|--------|
| Afrikaans | afr_Latn | Germanic | Latin |
| Swahili | swh_Latn | Bantu | Latin |
| Moroccan Arabic | ary_Arab | Semitic | Arabic |
| Somali | som_Latn | Cushitic | Latin |
| Amharic | amh_Ethi | Semitic | Ethiopic |
| Egyptian Arabic | arz_Arab | Semitic | Arabic |
| Hausa | hau_Latn | Chadic | Latin |
| Kinyarwanda | kin_Latn | Bantu | Latin |
| Zulu | zul_Latn | Bantu | Latin |
| Igbo | ibo_Latn | Volta-Niger | Latin |
| Plateau Malagasy | plt_Latn | Austronesian | Latin |
| Xhosa | xho_Latn | Bantu | Latin |
| Shona | sna_Latn | Bantu | Latin |
| Yoruba | yor_Latn | Volta-Niger | Latin |
| Nyanja | nya_Latn | Bantu | Latin |
| Southern Sotho | sot_Latn | Bantu | Latin |
| Tigrinya | tir_Ethi | Semitic | Ethiopic |
| Tunisian Arabic | aeb_Arab | Semitic | Arabic |
| Oromo | gaz_Latn | Cushitic | Latin |
| Tswana | tsn_Latn | Bantu | Latin |

**High-resource languages (for catastrophic forgetting mitigation):** English, French, Portuguese, Arabic

## Training Data

Our training corpus combines multiple high-quality sources:

- **African Monolingual Data** (~22.8B tokens): FineWeb2, WURA, and MADLAD-400
- **Code** (~1B tokens): CornStack-Python for reasoning capabilities
- **Mathematics** (~1B tokens): FineMath-4+ for mathematical understanding
- **Synthetic Data** (~324M tokens): GPT-4.1 translated domain-specific content across 10 domains
- **Parallel Data** (~456M tokens): NLLB-OPUS filtered with SSA-COMET (threshold 0.7)

We use **UniMax sampling** to create a balanced distribution, capping high-resource languages at approximately 1B tokens and upsampling lower-resource languages for up to five epochs.

## Quickstart

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "McGill-NLP/AfriqueQwen-8B"

# Load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

# Prepare the model input
prompt = "Bawo ni o ṣe n ṣe?"  # Yoruba: "How are you doing?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

# Generate text
generated_ids = model.generate(
    **inputs,
    max_new_tokens=100,
    do_sample=True,
    temperature=0.7,
    top_p=0.9
)
output = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
print(output)
```

## Deployment

For deployment, you can use `vllm` or `sglang` to create an OpenAI-compatible API endpoint:

**vLLM:**
```shell
vllm serve McGill-NLP/AfriqueQwen-8B
```

**SGLang:**
```shell
python -m sglang.launch_server --model-path McGill-NLP/AfriqueQwen-8B
```

## Training Details

### Hyperparameters

- **Learning Rate**: 5e-5 (with warmup and cosine decay)
- **Context Length**: 8,192 tokens (training)
- **Batch Size**: Effective batch size optimized for throughput
- **Optimizer**: AdamW
- **Precision**: BF16 mixed precision

### Infrastructure

Training was conducted using the LLaMA-Factory framework on up to 64 NVIDIA H100 GPUs with:
- DeepSpeed ZeRO-1/ZeRO-2
- Flash Attention 3
- Sequence packing
- Liger Kernel optimizations

## Evaluation

AfriqueQwen-8B is evaluated on multiple multilingual benchmarks including:

- **AfriMGSM**: Mathematical reasoning
- **AfriMMLU**: Multilingual knowledge
- **AfriXNLI**: Natural language inference
- **Belebele**: Reading comprehension
- **SIB-200**: Topic classification
- **FLORES**: Machine translation

## Model Variants

- [AfriqueQwen-14B](https://huggingface.co/McGill-NLP/AfriqueQwen-14B) - Larger, more capable variant

## Intended Use

This model is designed for:
- Research on African language NLP
- Building applications for African language communities
- Cross-lingual transfer learning experiments
- Multilingual text generation and understanding

## Limitations

- This is a **base/pre-trained model** and may require fine-tuning for specific tasks
- Performance varies across languages based on data availability
- May generate biased or inappropriate content without proper safeguards
- Not suitable for production use without additional safety measures

## Citation

If you find our work helpful, please cite:

```bibtex
@article{afriquellm2025,
  title={AfriqueLLM: How Data Mixing and Model Architecture Impact Continued Pre-training for African Languages},
  author={AfriqueLLM Team},
  year={2025}
}
```

## License

This model is released under the [Apache 2.0 License](https://www.apache.org/licenses/LICENSE-2.0). Please review the license terms before use.

## Acknowledgments

We thank the creators of the base models and datasets that made this work possible, including Alibaba (Qwen), the FineWeb team, WURA, MADLAD-400, and the NLLB project.