Abstract: Ternary large language models (LLMs), which use ternary precision weights and 8-bit activations, have demonstrated competitive performance while significantly reducing the high computational ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results