2024-02-29 15:39:47.AIbase.6.1k
Microsoft's 6-Page Paper Goes Viral: Ternary LLM, Truly Amazing!
The large language model enters the 1-bit era as Microsoft and University of Chinese Academy of Sciences propose the BitNet b1.58 method, which converts parameters to ternary representation, reducing model memory usage and simplifying the computation process. Performance optimizations enhance model speed and reduce memory consumption, sparking heated discussions and debates among netizens.
