使用rkllm工具链模型转换相关参数配置

#1

Screenshot%20from%202024-12-11%2015-12-35
Screenshot from 2024-12-11 15-12-35.png946×46 11.6 KB
我在转换模型时修改参数如上两图，其他代码与官网demo一致，目前转换的Qwen2-1.5B-Instruct模型，token数在3576上最多9.22token/s，官网相同的配置是14.40，请问还有什么因素影响呢？

#2

@lala

  • rkllm performance data were collected based on the maximum CPU and NPU frequencies of each platform with version 1.1.0.