New paper proposes adaptive expert offloading for multi-GPU MoE inference
An arXiv paper (2611.10219) addresses GPU memory constraints in Mixture-of-Experts inference by using expert offloading, which stores part of the model elsewhere, with an adaptive multi-GPU approach.
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-10 05:00 · arXiv cs.AI
Multi-GPU MoE Inference with Adaptive Expert Offloading