---
license: apache-2.0
language:
- en
- fr
- es
- pt
- sw
- ja
- ar
base_model: Qwen/Qwen2-0.5B-Instruct
tags:
- agriculture
- multilingual
- chatbot
- crop-diseases
- farming
- west-africa
pipeline_tag: text-generation
---

# AgriChat Multilingual - Agricultural Assistant

A multilingual chatbot fine-tuned for agricultural assistance, specifically designed for farmers in West Africa and beyond.

## Model Description

- **Base Model:** Qwen/Qwen2-0.5B-Instruct
- **Fine-tuning Method:** LoRA (Low-Rank Adaptation)
- **Languages:** English, French, Spanish, Portuguese, Swahili, Japanese, Arabic
- **Domain:** Agricultural crop diseases, farming practices, pest management
- **License:** Apache 2.0

## Supported Languages

| Language | Code | Coverage |
|----------|------|----------|
| English | en | Full |
| French | fr | Full |
| Spanish | es | Full |
| Portuguese | pt | Full |
| Swahili | sw | Full |
| Japanese | ja | Full |
| Arabic | ar | Full |

## Use Cases

- **Crop Disease Identification:** Ask about symptoms and treatments for plant diseases
- **Farming Advice:** Get guidance on agricultural practices
- **Pest Management:** Learn about controlling pests affecting crops
- **Multilingual Support:** Communicate in 7 different languages

## Quick Start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model
model = AutoModelForCausalLM.from_pretrained("mesabo/agri-chat-multilingual")
tokenizer = AutoTokenizer.from_pretrained("mesabo/agri-chat-multilingual")

# Chat example
messages = [
    {"role": "user", "content": "How do I identify cassava mosaic disease?"}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt")

outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
```

## Training Details

- **Training Data:** 38 curated Q&A examples across 7 languages
- **Epochs:** 3
- **LoRA Parameters:** 2.16M trainable (0.44% of total)
- **Training Loss:** 2.54
- **Hardware:** NVIDIA RTX 3090 (25.3 GB)
- **Training Time:** ~17 seconds

## Covered Topics

### Crop Diseases
- Cassava mosaic disease
- Maize leaf blight
- Tomato bacterial wilt
- Cashew anthracnose
- Rice blast disease

### Farming Practices
- Organic pest control
- Soil health management
- Crop rotation benefits
- Water conservation

## Limitations

- Fine-tuned on limited agricultural domain data
- Best suited for common crop diseases in West Africa
- May not cover specialized or rare conditions
- Responses should be verified with local agricultural experts

## Intended Use

This model is designed for:
- Agricultural extension workers
- Small-scale farmers
- Agricultural education platforms
- Farming assistance applications

## Citation

```bibtex
@misc{agri-chat-multilingual,
  author = {mesabo},
  title = {AgriChat Multilingual - Agricultural Assistant},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/mesabo/agri-chat-multilingual}
}
```

## Related Models

- [mesabo/agri-plant-disease-resnet50](https://huggingface.co/mesabo/agri-plant-disease-resnet50) - Plant disease image classification (95%+ accuracy)

## Contact

For questions or issues, please open a discussion on the model page.