BiRen Technology cooperates with Wuwenxinqiong to double the performance of large model training

BiRen Technology and Wuwen Xinqiong jointly developed and successfully implemented elastic fault tolerance, abnormal node detection and other functions on the Qianka training cluster, and improved the LLM model training performance by nearly 100%. The BiLi 110E inference card can reach up to 1.3 times the AI ​​computing power density of the mainstream cloud PCIE8 card server solution in the market, while saving 70% of energy consumption. The BiLi 110E inference card has been launched on Wuwen Xinqiong's large model service platform, supporting a variety of artificial intelligence inference application scenarios.