bakarqureshi/FoodExtract-gemma-3-270m-fine-tune-v1
The bakarqureshi/FoodExtract-gemma-3-270m-fine-tune-v1 is a 0.3 billion parameter language model built on Google's Gemma 3 270M architecture, specifically fine-tuned for food and drink extraction from raw text. It classifies text as food/drink, tags it with predefined categories, and extracts lists of food and drink items. 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 bakarqureshi, is a specialized language model based on the Gemma 3 270M architecture, fine-tuned for the precise extraction of food and drink information from raw text. With 0.3 billion parameters and a context length of 32768 tokens, it's designed for efficiency in specific information retrieval tasks.
Key Capabilities
- Food/Drink Classification: Determines if input text contains food or drink references.
- Tagging: Assigns relevant tags (e.g., 'food_items', 'drink_items', 'ingredient list', 'menu') to the text for quick categorization.
- Item Extraction: Identifies and lists specific food and drink items mentioned in the text.
- Optimized Output: Generates condensed, structured outputs (not raw JSON) to minimize token generation, with helper functions provided for easy conversion to and from a more detailed dictionary format.
Training and Use Cases
The model was trained using Supervised Fine-Tuning (SFT) on the FoodExtract-1k dataset, which comprises 1400 samples of raw text and structured food extractions generated by gpt-oss-120b. This makes it particularly effective for:
- Data Filtering: Efficiently filtering large text datasets, such as image captions (e.g., DataComp-1B), to isolate food and drink-related entries.
- Content Analysis: Structuring unstructured text data related to culinary topics for further analysis or database population.