This repository is for Artifact Evaluation of the International Symposium on Computer Architecture (ISCA) 2024 paper MAD Max Beyond Single-Node: Enabling Large Machine Learning Model Acceleration on Distributed Systems.
The two main folders for reproducing the performance model results in the paper are:
-
artfiact_notebooks- Jupyter Notebooks used for launching experiments to recreate performance model results[0] Cloud Provider Launcher.ipynb- used for recreating Figures 1 and 16[1] DLRM A Validation.ipynb- used for recreating Figure 7
-
artifact_sheets- Microsoft Excel sheets that contain organized data of experiment results[A] Cloud Provider Results.xlsx- supplemental data used in Figures 1 and 16
The other files are used to support the performance model itself. <More to come on description of files' structure.>
Please note that these two folders are still works in progress and will updated over the next couple of days.