---
library_name: transformers
language:
- wo
- en
license: apache-2.0
---

# Oolel: Pioneering Open Source Wolof Language Model

<!-- Provide a quick summary of what the model is/does. -->


## 1. Model Description

<!-- Provide a longer summary of what this model is. -->


**Oolel** is Africa's first open-source language model designed specifically for **Wolof**, one of West Africa's major languages. Created by **Soyno**, it introduces cutting-edge language technology to Wolof speakers, making powerful language technology accessible to everyone.

Based on **Qwen 2.5** architecture, Oolel brings together cutting-edge AI advances and Wolof linguistic expertise. This pioneering work stands as a testament to Soyno's commitment to developing AI solutions by Africans, for Africa, marking a significant step toward the continent's technological sovereignty.

- **Developed by: Soynade Research (Soyno)**
- **Supported Languages:**  Wolof, English, French
- **Status:** This is a static model trained on an offline dataset. Future versions may be released that improve model capabilities.
- **Model Release Date:** Dec 04, 2024
- **License:** Apache License
- **Finetuned from model:** Qwen 2.5

## 2. Key Features and Capabilities
Oolel demonstrates proficiency in:

- Bidirectional translation between English and Wolof
- Natural text generation in Wolof
- Code generation with Wolof instructions
- Standard LLM tasks including:
  - Question answering
  - Summarization
  - Contextual understanding

## 3. Usage

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

### 3.1 With Transformers
```python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

device = "cuda" 

model = AutoModelForCausalLM.from_pretrained(
    "Soyno/Oolel-v0.1",
    torch_dtype = torch.bfloat16,
    device_map="auto")

tokenizer = AutoTokenizer.from_pretrained("Soyno/Oolel-v0.1")

def generate_response(messages, max_new_tokens=1024, temperature=0.1):
    text = tokenizer.apply_chat_template(
          messages,
          tokenize=False,
          add_generation_prompt=True
)
    model_inputs = tokenizer([text], return_tensors="pt").to(device)
    generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=max_new_tokens, temperature=temperature)
    
    generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
    return response

```


**Some tasks examples:**

1. **Translation Tasks**

```bash
messages = [
    {"role": "user", "content": "Translate to Wolof: Bassirou Diomaye Faye is the new Senegalese president. He is 44 years old"}
]
print(generate_response(messages))
```

2. **Code generation**

```python
messages = [
    {"role": "user", "content": "Bindal ab klaas Python buy wone ni ñuy jëfandikoo dataframe yi ci Pandas"}
]
print(generate_response(messages))
```

3. **Problem Solving**

```bash
system_prompt = "You're a Wolof AI assistant. Please always provide detailed and useful answers to the user queries"
messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "Ndax nga mën ma won ni ñuy resolver problème bii: Fatou dafa jënd 3 kilo ceeb, 2 kilo diw ak 5 kilo sukër. Ceeb gi wenn kilo 500 CFA la, diw gi 1200 CFA kilo bi, sukër gi 750 CFA kilo bi. Ñaata la wara fay?"}
]
from pprint import pprint
pprint(generate_response(messages))
```

  
4.  **Text Generation** (e.g. story generation)

```python
system_prompt = "You are a skilled Wolof storyteller (Gewël) with deep knowledge of African folktales and traditions. Write engaging stories in Wolof that reflect African cultural values and wisdom."

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "Bindal ab léeb ci gaynde gi lekk muus mi"}
]
print(generate_response(messages, temperature=0.9))
```

5. **Multi-turn conversations**

```bash
messages = [
   {"role": "user", "content": "Wax ma clan mooy CEDEAO ? Ci lan la liggeey?"},
   {"role": "assistant", "content": "CEDEAO mooy 'organisation' gu boole reew yi nekk ci pennc Afrika bi. Mu ngi sukkandiku ci wàll économie, politig, ak déggoo diggante reew yi"},
   {"role": "user", "content": "ñaata reew ñoo ci bokk?"}
]
print(generate_response(messages))
```

### 3.2 VLLM

```python
from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

tokenizer = AutoTokenizer.from_pretrained("Soyno/Oolel-v0.1-Instruct")

# Pass the decoding hyperparameters
sampling_params = SamplingParams(temperature=0.7, top_p=0.8, repetition_penalty=1.05, max_tokens=512)

llm = LLM(model="Soyno/Oolel-v0.1-Instruct")

prompt = "Kan mooy Youssou Ndour ?"
messages = [
    {"role": "system", "content": "You are Oolel, created by Soyno. You are a helpful assistant. Please always provide detailed and useful answers to the user queries"},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

outputs = llm.generate([text], sampling_params)

# Print the outputs.
for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt}, Generated text: {generated_text}")
```

## 4. Bias, Risks, and Limitations

While Oolel marks a significant milestone, we acknowledge its current limitations:

- As a first version, the model's performance continues to evolve
- Training data diversity can be further expanded
- Specific domain expertise can be enhanced

Future developments will focus on:

- Enriching training data with comprehensive African historical content
- Deeper integration of cultural contexts and nuances
- Improving performance across various linguistic tasks
- Strengthening the model's ability to handle complex cultural contexts


## 5. Authors
- **Yaya SY**: NLP Researcher (Efficient Continued Pretraining)
- **Dioula DOUCOURE**: Data & NLP Engineer