walterwhite91/ask-m-gemma3-guide-lora

Ask-M Gemma-3 Guided Learning LoRA

This repository contains a LoRA adapter fine-tuned for guided, step-by-step educational explanations using the Gemma-3 12B model.

The adapter focuses on teaching concepts clearly, not just answering questions.

---

🎯 Purpose

This LoRA is optimized to:

• Explain concepts step-by-step
• Use simple, student-friendly language
• Structure answers logically
• Emphasize understanding over brevity

It complements the exam-focused adapter by supporting learning and revision workflows.

---

🧠 Training Overview

• Base model: unsloth/gemma-3-12b-it-bnb-4bit
• Fine-tuning method: QLoRA
• Trainable parameters: ~24M ~0.2% of base model)
• Sequence length: 1024
• Epochs: 2
• Optimizer: paged_adamw_8bit
• Hardware: NVIDIA T4 (Kaggle)

---

📂 Dataset Format

Each example follows this format:

{
  "instruction": "You are a tutor. Explain the concept clearly...",
  "input": "Subject: PHYS101\nQuestion: ...",
  "output": "Step-by-step explanation...",
  "meta": {
    "subject": "PHYS101",
    "keywords": ["..."]
  }
}