Simple Training Skeleton for VLMs

This repository contains code to quickly train visual-language models from pre-trained models (e.g., ~300 lines of code for model, dataset, and training).

Supported Vision Models

• openai/clip-vit-base-patch32
• google/siglip-base-patch16-224

Supported Language Models

• allenai/OLMo-1B-hf
• google/gemma-2b-it
• mistralai/Mistral-7B-Instruct-v0.3