taozhiyuai

open as many terminals as you can

> > open as many terminals as you can > > I ended up typing this command: > > ``` > ollama pull llava-phi3; ollama pull llava-llama3; ollama pull llama3-gradient;...

is it the right model file? or I should merge them into one GGUF? FROM ./Meta-Llama-3-120B-Instruct-Q5_K_M-00001-of-00003.gguf FROM ./Meta-Llama-3-120B-Instruct-Q5_K_M-00002-of-00003.gguf FROM ./Meta-Llama-3-120B-Instruct-Q5_K_M-00003-of-00003.gguf FROM ./Meta-Llama-3-120B-Instruct.imatrix TEMPLATE """{{ if .System }}system {{ .System }}{{...

已经处理过这个问题,搜索一下

he may ask to keep model in GPU RAM all the time, even it is not in use for a very long time.

> > he may ask to keep model in GPU RAM all the time, even it does not used for a very long time. > > Yes, that's what it...

多的是中文版,搜索一下

deepseek team does not offer gguf, which in not in their plan at the moment. and its performance is not much better than llama-3-70b, but 3 times bigger in parameters...

@RexWzh @jmorganca for the same model or quantizations , is that possible the website combine them into one link. so difference users can contribute together.

Last login: Tue May 7 07:23:10 on ttys001 taozhiyu@603e5f4a42f1 Q5KM % ollama create meta-llama-3-120b-instruct:q5_k_m -f modelfile transferring model data creating model layer creating template layer creating parameters layer creating config...