You can deploy this model and start earning money today!

diegoluchetti1/
nasa-se-assistant-3b-gguf

Name: nasa-se-assistant-3b-gguf
Availability: InStock

gguf

endpoints_compatible

conversational

Views

No views yet

NASA SE Assistant 3B (GGUF)

A small language model fine-tuned for NASA Software Engineering and embedded systems. It answers questions using NASA terminology (e.g. NPR 7150.2, Verification & Validation, requirements “Shall”) and is intended for local inference.

Model details

Base model: Qwen 2.5 3B Instruct
Fine-tuning: LoRA (SFT) on a synthetic NASA SE / embedded-systems QA dataset
Format: GGUF, F16 (~6.4 GB)
Use case: RAG-style assistant over NASA SE docs; local chat or API

Files

nasa_se_assistant_3b.F16.gguf — full model in GGUF F16

How to use

Ollama

Create a Modelfile with:
FROM /path/to/nasa_se_assistant_3b.F16.gguf
Run:
bash
1ollama create nasa-se-assistant -f Modelfile 2ollama run nasa-se-assistant

llama.cpp

./llama-cli -m nasa_se_assistant_3b.F16.gguf -p "Your prompt" -n 256

Python (with Ollama API)

With Ollama running and the model loaded:

python

import requests
r = requests.post(
    "http://localhost:11434/api/generate",
    json={"model": "nasa-se-assistant", "prompt": "What is NPR 7150.2?"}
)

Training

Data: Synthetic instruction/response pairs (NASA SE, requirements, V&V).
Method: Unsloth + LoRA; base Qwen 2.5 3B.
Hardware: Single GPU (e.g. Tesla T4).

License

Refer to the base model (Qwen 2.5) license. This fine-tune is for educational / internal use.

Citation

If you use this model, please cite the base model (Qwen 2.5) and this project (NASA SE Assistant).

diegoluchetti1/nasa-se-assistant-3b-gguf