Reproducibility Report for ACM SIGMOD 2024 Paper: 'CAFE: Towards Compact, Adaptive, and Fast Embedding for Large-scale Recommendation Models'

We have been able to reproduce the original paper's findings partially. The authors provided the documentation with clear instructions on reproducing each experiment, from pre-processing to training. While most of the code can be run without any modifications, some scripts require manual changes to select particular datasets. Moreover, some experiments failed with an error or were not reproduced due to hardware limitations. All experiments with avazu and criteo datasets were reproduced. Additionally, HotSketch performance tests were successfully repeated.