FastVLM

May 13, 2011 - Apple's machine learning team released and open-sourced a visual language model, FastVLM, on GitHub last week, with versions 0.5B, 1.5B, and 7B. According to the introduction, the model is developed based on Apple's own MLX framework and trained with the help of LLaVA codebase, which is optimized for end-side AI operations of Apple Silicon devices. According to the technical document, FastVLM realizes near real-time response for high-resolution image processing while maintaining accuracy, and the required counting...