Gifted Maths LoRA Adapter
This repository contains the separated LoRA adapter for maths_engine, a local gifted mathematics reasoning model.
Base model: unsloth/gemma-4-E2B-it
Training goal:
- step-by-step mathematical reasoning
- readable LaTeX formatting
- local/offline deployment after GGUF conversion
Important:
- This is the adapter only.
- To create a full local model, load it with the base model and merge using Unsloth
save_pretrained_merged(..., save_method="merged_16bit").
- Do not rely on ordinary
PeftModel.merge_and_unload() for this Gemma 4 workflow.
