The web content showcases the performance of deepseek_r1_distill_llama_8b_q40 model on different configurations of Raspberry Pi 5 8GB. Surprisingly, the evaluation prediction for 2 x Raspberry Pi is 7.70 tok/s, while for 4 x Raspberry Pi it is 11.68 tok/s. The unique aspect of this content is the detailed breakdown of processing and memory usage for each configuration. However, the controversial aspect may be the comparison of performance between different numbers of Raspberry Pi devices. Overall, the content provides valuable insights into the performance metrics of the model on Raspberry Pi devices.
https://github.com/b4rtaz/distributed-llama/discussions/162