README.md

---
title: GeoLLM - Mineral Exploration Geology
emoji: ⛏️
colorFrom: yellow
colorTo: green
sdk: gradio
sdk_version: 5.12.0
app_file: app.py
pinned: false
license: apache-2.0
models:
  - AshkanTaghipour/GeoLLM-Qwen3.5-0.8B
datasets:
  - AshkanTaghipour/mineral-exploration-geology-qa
---

# GeoLLM — Mineral Exploration Geology Assistant

An interactive demo of **GeoLLM-Qwen3.5-0.8B**, a domain-adapted LLM for mineral exploration geology.

Ask questions about deposit models, geochemistry, geophysics, drilling strategies, and Western Australian geology — the model answers like a knowledgeable colleague, concise and technically grounded.

## Model Details

- **Base model:** [Qwen3.5-0.8B](https://huggingface.co/Qwen/Qwen3.5-0.8B)
- **Fine-tuning:** bf16 LoRA (r=16) on 479 expert QA pairs from Western Australian exploration reports
- **Format:** float32 (~3.4 GB) on CPU via HuggingFace transformers
- **Inference backend:** [transformers](https://github.com/huggingface/transformers) with streaming

## Links

- **All models (0.8B–27B):** [HuggingFace](https://huggingface.co/AshkanTaghipour)
- **Training code & benchmarks:** [GitHub](https://github.com/AshkanTaghipour/GeoLLM-Qwen3.5-FineTune)
- **Dataset:** [mineral-exploration-geology-qa](https://huggingface.co/datasets/AshkanTaghipour/mineral-exploration-geology-qa)