Empowering Domestic Chips: Overcoming FP8 Model Deployment Challenges and DeepSeek Costs, Tsinghua Team Opens Source Large Model Inference Engine 'Chitu'
2025-03-14 / Read about 0 minute
Author:小编   

The team led by Professor Zhai Jidong from the Institute for High Performance Computing at Tsinghua University, in partnership with Qingcheng Jizhi, a Tsinghua-affiliated science and technology innovation enterprise, has jointly announced the open-source release of the large model inference engine 'Chitu'. This groundbreaking engine is the first to natively support FP8 precision models on non-NVIDIA Hopper architecture GPUs and various domestic chips, heralding a significant advancement for the domestic AI chip ecosystem. This achievement not only lowers the barriers to deploying large models but also accelerates the development of domestic AI computing capabilities.