List your company

Deduplication on RealNews 0.1M documents

7,158Duplication RatioGPU baseline
7,151.647,194.577,237.57,280.43Jan 2, 2025

Evaluation results

MethodLinks
GPU baselineImplementation=NVIDIA NeMo Curator, Hardware=4 GPUs (V100)2025.017,15895.2
SEDDHardware=4 GPUs (V100)2025.017,17695.6
CPU baselineImplementation=SlimPajama, Hardware=64 logical CPU cores2025.017,28499.5
Standard MinHashType=Oracle/Brute-force2025.017,317—