Milestone: Ascend 910C Completes Full-Parameter Post-Training of 1.6 Trillion Parameter Model, Domestic AI Computing Crosses Key Threshold
Shenzhen Hetao College, in collaboration with Harbin Institute of Technology (Shenzhen), the Shenzhen Big Data Research Institute, and Huawei, has achieved a landmark breakthrough in domestic AI computing. The consortium successfully completed full-parameter post-training of the 1.6 trillion-parameter DeepSeek-V4-Pro model using the Ascend 910C domestic AI computing cluster — marking the first time China's homegrown chips have powered a model at this scale through the entire
A collaborative effort involving Shenzhen Hetao College, Harbin Institute of Technology (Shenzhen), the Shenzhen Big Data Research Institute, and Huawei has successfully completed full-parameter post-training for the 1.6 trillion-parameter DeepSeek-V4-Pro model. This significant achievement was powered by the Ascend 910C domestic AI computing cluster, marking the first time Chinese-made chips have supported a model of this magnitude throughout its entire training process. The project overcame substantial technical hurdles, including managing massive data exchange and ensuring system stability during the training of the complex Mixture-of-Experts architecture. The team implemented novel solutions for parameter distribution, load balancing, and fault tolerance, resulting in high computing utilization and improved efficiency.
This milestone demonstrates that China's indigenous AI chips are now capable of supporting world-class large model training, a critical development for the nation's technological self-sufficiency and advancement in artificial intelligence amidst international semiconductor restrictions.
📌 Kaynak
Bu özet Pandaily kaynağından otomatik derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →