mrdbourke/FoodExtract-gemma-3-270m-fine-tune-v1
The mrdbourke/FoodExtract-gemma-3-270m-fine-tune-v1 is a 0.3 billion parameter language model, fine-tuned from Google's Gemma 3 270M, specifically designed for food and drink extraction from raw text. It classifies text as food or drink, tags it with relevant categories, and extracts edible food and drink items into structured lists. This model is optimized for filtering large text datasets, such as image captions, to identify food and drink-related content.
Loading preview...
FoodExtract-gemma-3-270m-fine-tune-v1 Overview
This model, developed by mrdbourke, is a specialized language model built upon the Gemma 3 270M architecture. Its core function is to process raw text and extract structured information related to food and drink. The model performs three key tasks:
Key Capabilities
- Classification: Determines if input text describes food or drink items.
- Tagging: Assigns one or more predefined tags (e.g., 'food items', 'drink items', 'ingredient list', 'recipe') to the text for quick filtering.
- Extraction: Identifies and lists specific edible food and drink items mentioned in the text.
Training and Data
The model was fine-tuned using Supervised Fine-Tuning (SFT) with Hugging Face's TRL library. It was trained on the FoodExtract-1k dataset, which comprises 1400 samples of raw text and corresponding structured food extraction outputs generated by gpt-oss-120b. The model is designed to output a condensed, token-efficient format, with helper functions provided for converting between condensed and uncondensed structured data.
Good For
- Data Filtering: Ideal for filtering extensive text datasets, such as image captions (e.g., DataComp-1B), to isolate food and drink-related entries.
- Structured Data Generation: Useful for transforming unstructured text into structured lists of food and drink items, along with relevant classifications and tags.