Search papers, labs, and topics across Lattice.
Affiliation:
2
0
1
Decentralized robotics data can yield stronger manipulation policies than centralized training: slotting locally trained client LoRAs directly into a server MoE beats centralized PEFT on real hardware by 12.23% while slashing communication by 87%.
This paper presents ETCInfer, an energy-efficient, thermal-aware scheduler that selects a pre-job Computer Room Air Conditioner (CRAC) setpoint and adapts per-GPU frequency and micro-batch size during execution and designs ETCAdapter, a learning-based controller that minimizes per-job energy under thermal safety and SLO constraints.
