The Offline Dilemma!
Physical textbooks lack active retrieval! WhyBook is powered by the incredibly fast gemma-4-e2b model fine-tuned to give students the 'Why' at light-speed!
The Payload Schema!
WHAT?
The core concept: A dynamic knowledge graph that adapts to student inquiry in real-time. No more static text!
WHY?
Pedagogical active retrieval. By forcing the "Why", we bypass rote memorization and build deep neural connections.
REAL WORLD
From classroom labs to garage builds—connecting abstract physics to the tangible tools in your hands.
The Multiverse Pipeline!
Environment Boot!
DANGER: Kaggle T4 15GB VRAM Limit! OPTIMIZE OR PERISH!
Synthetic Yield!
90 Curated QA Pairs generated via OpenAI JSON schema injection for maximum precision.
Unsloth Training!
- ⚡ gemma-4-e2b QLoRA
- ⚡ Rank 16 / Alpha 32
- ⚡ Peak VRAM: 12.1 GB
BEATING OOM CRASHES!
GGUF Export!
Quantization to 4-bit precision for lightning-fast edge inference.
Payload Artifacts
Top Secret Files!
README.md
Project overview, workflow summary, and direct links to the published model, dataset, and Space.
Kaggle Notebook
The Phase 2 fine-tuning notebook for Gemma 4 E2B, Unsloth, QLoRA, and GGUF export.
HF Space
Published Gradio Space files for the shareable WhyBook chemistry tutor demo.