Ternair
Run a 4B-parameter LLM in 942 MiB — PyTorch-native BitNet b1.58 with zero float multiplications at inference
原文来源AI 开源 Releases
查看官方原文
Run a 4B-parameter LLM in 942 MiB — PyTorch-native BitNet b1.58 with zero float multiplications at inference