Supplemental materials to "Benchmarking Local Language Models for Social Educational Robots using Edge Devices (Raspberry Pi4)"

This place contains all the results from the benchmarking of the master thesis : benchmarking and deploying local language models for social educational robots using edge devices.All the codes and explanation to run the benchmark are available on : https://github.com/RobotStudyCompanion/Benchmarking_LLM.gitThis work was about benchmarking 25 open sources Language model on the RaspberryPi 4 model B (8GB RAM) on 8 metrics (e.g : Token Per Second, Token Per Joule, Teaching Effectiveness, MMLU Score).The results are stored in the .json and .csv files.A lot of graph has been also uploaded to give a better representation of the different model performance during the benchmark (.jpg file)if you have any question concerning the work or the results you can contact me at : dorianlamouille@gmail.com