A principled approach for energy-efficient training via phase-aware GPU frequency tuning
Modern AI model training imposes unprecedented computational demands, making it a key contributor to datacenter energy consumption. Yet a significant fraction of the energy consumed during training does not translate to useful computation due to bottlenecks throughout the training pipeline. We present PAFT, a phase-awa...
Miguel Braga, Júlio Pinto, Rahma Nouaji et al.
· 0 citations