Reproducing EL-Rec: Efficient Large-scale Recommendation Model Training via Tensor-train Embedding Table

I cloned this repository from the artifact evaluation repository for the recommendation model proposed in EL-Rec: Efficient Large-scale Recommendation Model Training via Tensor-train Embedding Table. It includes updates I made to the scripts to allow them to run successfully in addition to some preliminary work to train the model on a new dataset. To reproduce the results, I used 4 V100 16GB NVLink GPUs running Ubuntu 20.04, containing 8 vCPUs, 24 GB RAM, and a shared SSD of 700GB. My results and methodology are explained in detail in the paper Reproducing-El-Rec.

Note: A few of the code changes use hardcoding or are otherwise just quick fixes that were necessary to get the scripts working by the end of the semester (hence the lack of a PR to the original repository). They do not represent the intent to propose a modification to the original work.