README.md

metadata

license: mit
library_name: transformers
tags:
  - nasa
  - space-science
  - research-assistant
  - phi3
  - rag
  - ollama
language:
  - en
pipeline_tag: text-generation

NASA Research Assistant (Ollama Model)

Model Description

This is a specialized language model trained on NASA research data, designed to assist with space science questions, research summarization, and technical explanations. The model is based on Phi-3 Mini 3.8B and has been optimized for scientific accuracy and detailed explanations.

Model Details

• Base Model: Microsoft Phi-3 Mini 3.8B
• Training Data: NASA research papers, mission reports, and scientific publications
• Model Size: ~3.8B parameters
• License: MIT
• Format: GGUF (compatible with Ollama, llama.cpp)

Training Data Sources

• NASA research papers from PubMed Central (3,938 documents processed)
• NASA Taskbook project descriptions
• NASA Open Science Data Repository
• Space biology and microgravity research
• Astronaut health and safety studies
• Planetary science publications

Capabilities

• Question Answering: Detailed responses about space science topics
• Research Summarization: Condensing complex scientific papers
• Technical Explanations: Breaking down aerospace concepts
• Mission Analysis: Discussing NASA missions and findings
• Scientific Accuracy: Trained on peer-reviewed research

Usage

With Ollama

# Pull the model
ollama pull Ravurijeetendra12/ollama-nasa-model

# Run the model
ollama run Ravurijeetendra12/ollama-nasa-model

Example Queries

• "What are the main health risks for astronauts on long-duration missions?"
• "Explain the effects of microgravity on plant growth"
• "Summarize recent findings about space radiation exposure"
• "How does the ISS maintain its orbit?"

Model Performance

The model has been optimized for:

• Scientific accuracy in space-related topics
• Detailed explanations suitable for researchers and students
• Proper citation of NASA research when applicable
• Clear communication of complex concepts

Limitations

• Knowledge cutoff based on training data (up to 2024)
• Primarily focused on NASA and US space research
• May not have complete coverage of very recent developments
• Should not be used for mission-critical decisions without verification

Training Process

1. Data Collection: Gathered NASA research papers and publications
2. Data Processing: Cleaned and chunked scientific texts
3. RAG Integration: Built retrieval-augmented generation system
4. Fine-tuning: Applied LoRA fine-tuning on instruction pairs
5. Optimization: Configured for scientific accuracy and detail

Technical Specifications

• Context Length: 4096 tokens
• Temperature: 0.7 (balanced creativity/accuracy)
• Top-p: 0.9
• Quantization: Q4_K_M (recommended for most use cases)

Citation

If you use this model in your research, please cite:

@misc{nasa-ollama-model-2024,
  title={NASA Research Assistant: Specialized Language Model for Space Science},
  author={Ravurijeetendra12},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/Ravurijeetendra12/ollama-nasa-model}
}

Contact

For questions about this model or to report issues, please open an issue on the repository.

Acknowledgments

• NASA for providing open access to research data
• Microsoft for the Phi-3 base model
• The open-source community for tools and libraries

---

This model is designed for educational and research purposes. Always verify critical information with official NASA sources.