This repository is for the code portion of Green Lab course from VU Amsterdam. The code portion R scripts for data analysis. Prompt.docs the prompts used for the three interactions types used in the experiment.
To check normality of our data you can run Normality.R script in the root directory. It will generate histograms, qq-plots and timeline distribution of value per Machine-Task-LLM. Moreover it will generate this visualisations for both cpu and power usage.
The First Research Question Hypothesis is tested inside of Power-Kruskal-Wallis-test.R script in the root directory. When you run this script it will print the values for all the tests and their corresponding p-value.
If you want to check the power consumption of the experiment per machine per llm per task, you can run Power-usage-Watts.R script in the root directory.
To recreate the data visualizations go to ./data/dataVisualization.R. All the generated visuals are placed in ./data/visualization.
Interaction script with Selenium and RunnerConfig file for experiment-runner.