DeepSeek releases Janus Pro, with Intel Gaudi 2D aiding optimization.

Recently, DeepSeek officially launched its latest AIGC (Artificial Intelligence Generated Content) model—Janus Pro. This model has quickly captured industry attention with its high precision, outstanding performance, and optimized generation capabilities.

Janus Pro: Breakthrough Unified Architecture, Enhancing Text-to-Image Capability

Janus Pro is the first to adopt a unified Transformer architecture, overcoming the limitations of traditional AIGC models that rely on multi-path visual encoding, enabling it to support both understanding and generation tasks simultaneously. With optimized training strategies and an expanded model scale, Janus Pro demonstrates greater stability and performance enhancements in text generation and text-to-image tasks.

Currently, Janus Pro offers 1 billion and 7 billion parameter versions, further enhancing its adaptability across different application scenarios, providing the industry with more accurate and efficient AI content generation solutions.

Intel Gaudi 2D AI Accelerator Optimization, Enhancing Computational Efficiency

On February 2, Intel announced that the Intel® Gaudi 2D AI Accelerator had been optimized for Janus Pro. This accelerator, with its high bandwidth (2.45TB/s) and 96GB of HBM large-capacity memory, significantly boosts the throughput of text-to-image tasks, reducing dependence on computational hardware capability.

With optimizations in combination with the Intel Optimum-Habana framework, Janus Pro can generate 16 high-quality images in only about 10 seconds, greatly enhancing task processing efficiency and lowering deployment costs compared to traditional models.

AI Ecosystem Development: Promoting Technology Adoption

In addition to Janus Pro, DeepSeek's distillation models have also received support for the Xeon and Gaudi platforms, ensuring high precision while reducing energy consumption and operational costs (TCO).

Currently, Intel has provided optimization support for over 500 open-source AI models (including DeepSeek, Llama 3.1, Tongyi Qianwen, etc.) and continues to advance the construction of an open AI ecosystem to facilitate the deployment and promotion of enterprise-level AI applications.

With the launch of Janus Pro and the optimization of the Intel Gaudi 2D AI Accelerator, computing efficiency and content generation capabilities in the AIGC field are achieving new breakthroughs, bringing more possibilities to the AI industry.

The market carries risks, and investment should be cautious. This article does not constitute personal investment advice and has not taken into account individual users' specific investment goals, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investing based on this is at one's own responsibility.

Very good!（633）