Microsoft’s DeepSpeed-VisualChat: Breaking Boundaries in Multi-Modal Language Models

In a new paper DeepSpeed-VisualChat: Multi-Round Multi-Image Interleave Chat via Multi-Modal Causal Attention, a research team from DeepSpeed of Microsoft presents the DeepSpeed-VisualChat framework, which is designed to optimize LLMs by incorporating multi-modal capabilities, demonstrating superior scalability, even up to a 70 billion parameter model size.

ai artificial intelligence attention billion breaking breaking boundaries capabilities chat deep-neural-networks deepspeed framework image language language model language models llms machine learning machine learning & data science microsoft ml multi-modal paper research research team scalability team technology