README.md

metadata

base_model: google/gemma-3-4b-it
library_name: peft
license: gemma
datasets:
  - cindyliang/receipts-v1

Introduction

This fine-tuned model using Gemma 3 4B Instruct for receipts extraction.

Pre-processing Dataset

1. Check all receipt photos to see if are RGB mode.
2. Extract and parse "ground_truth" column.
3. Check if store_name, date and total have "" or not.
4. All store name must capital using upper()
5. Parsing date to "%d-%m-%Y" format
6. Check if total have $ or not, if total don't have $, add $