According to a report by Pinplay on October 19, the Arxiv page shows that Microsoft Research, together with the University of Chinese Academy of Sciences and Tsinghua University, jointly released a paper and announced a scalable and stable 1-$BIT$ Transformer architecture BitNet architecture. BitNet is designed for large language models. The research team said that to train the 1-bit weight, the team introduced BitLinear as nn. An alternative to the $LINEAR$ layer. Experimental results show that BitNet achieves competitive performance in language modeling tasks while significantly reducing memory footprint and energy consumption. In addition, BitNet presents a law of scaling similar to that of full-precision Transformers, showing that it has the potential to scale to larger language models while maintaining efficiency and performance benefits.

#bitcoin #microsoft

#axs #zkevm #etf #enjoy #crypto2023 #cryptonews