search

airockchip / rknn-llm

Other
417 stars 36 forks source link

benchmark data #9

Open WangFengtu1996 opened 7 months ago

WangFengtu1996 commented 7 months ago
fydeos-alex commented 7 months ago

The model benchmark can be found on the huggingface model card and here https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard

I think the model hardware request, text generate speed, and runtime cost are more important to test.

  • 能提供已经支持的LLM模型，在rk3588 上面的benchmark 数据么？
WangFengtu1996 commented 7 months ago

关于加载模型的内存使用量，token生成速率的数据

fydeos-alex commented 7 months ago

You're kidding, token speed and memory cost depend on various parameters you set, which are not typically included in benchmarks. I suggest you open a new issue and ask others to share their runtime data. That would be better.