food-vlm-tiny-quality-v2 (merged full model)

Merged full model checkpoint produced from local fine-tuning for food image to nutrition JSON generation.

Training data

• Dataset: Codatta/MM-Food-100K
• Training approach: supervised fine-tuning on conversational vision-language examples

Intended output format

The model is optimized for JSON-only responses with ingredient and nutrition structure.

Quickstart

from transformers import AutoProcessor, AutoModelForImageTextToText

repo = "thyfriendlyfox/food-vlm-tiny-quality-v2-merged"
processor = AutoProcessor.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained(repo, trust_remote_code=True)

Base model