Democratizing Access To Farming Knowledge
BackDropBuild Hackathon Repo for Agriculture Assistant by Caleb DeLeeuw.
Demo spaces and more full repo at: https://huggingface.co/spaces/Solshine/BDB-AgricultureAssistant/tree/main
Project Overview:
Tiny Llama Farmer aims to empower off-grid farmers with the knowledge of cutting-edge AI for organic, climate-mitigating, and regenerative agriculture. This project proposes a mobile app that leverages lightweight versions of TinyLlama LLM and TinyDolphin LLM, specifically optimized for on-device processing on modern smartphones.
Key Features:
- On-device AI: TinyLlama and TinyDolphin LLMs will be shrunk and fine-tuned using techniques like QLoRA and GLAN training, allowing them to run directly on mobile devices without an internet connection.
- Expert-curated Data: The fine-tuning process will utilize unique datasets created by agricultural experts, combining both human knowledge and synthetic data.
- Offline Assistance: Farmers in cellular-dead zones or off-grid locations can access crucial agricultural information and guidance.
- Regenerative Practices: The app will focus on promoting sustainable farming techniques that benefit the environment and local ecosystems.
Technical Approach:
- Model Shrinking: TinyLlama and TinyDolphin LLM architectures will be compressed using techniques like quantization and pruning to minimize their size and computational needs.
- On-device Training: Fine-tuning the models on expert-created and semi-synthetic datasets will be done directly on the phone, ensuring offline functionality.
- User Interface: A user-friendly mobile app will be developed to allow farmers to interact with the AI models for various tasks, such as:
- Crop disease identification and treatment recommendations
- Soil analysis and improvement suggestions
- Sustainable planting techniques and crop rotation plans
- Pest management strategies
Impact:
Tiny Llama Farmer has the potential to revolutionize agriculture in remote areas by:
- Boosting Food Production: Providing real-time, AI-powered assistance can improve crop yields and overall farm productivity.
- Promoting Sustainability: The focus on regenerative practices can lead to healthier soil, reduced environmental impact, and climate change mitigation.
- Empowering Farmers: Off-grid farmers gain access to valuable knowledge and resources, leading to greater self-sufficiency and economic opportunity.
Backdrop Build Hackathon Fit:
This project aligns perfectly with the Backdrop Build Hackathon's goals by promoting innovation and accessibility in technology. Tiny Llama Farmer addresses the need for offline solutions that empower individuals in underserved regions with cutting-edge tools for a more sustainable future.
Next Steps:
The project will require expertise in mobile app development, efficient AI model compression techniques, and agricultural data curation. By collaborating with relevant experts and leveraging the Backdrop Build Hackathon resources, Tiny Llama Farmer has the potential to bring transformative agricultural assistance directly into the hands of off-grid farmers.
Post Hackathon Notes Please see, on the Copyleft Cultivars Hugging Face Hub, the releases which followed up this project: https://huggingface.co/CopyleftCultivars
And especially to this model: https://huggingface.co/CopyleftCultivars/llama-3.1-natural-farmer-Q8_0-GGUF