Similar Items: Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference
- MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems
- Piper: Efficient Large-Scale MoE Training via Resource Modeling and Pipelined Hybrid Parallelism
- Relay Buffer Independent Communication over Pooled HBM for Efficient MoE Inference on Ascend
- Accelerating MoE with Dynamic In-Switch Computing on Multi-GPUs
- Space Network of Experts: Architecture and Expert Placement
- FaaSMoE: A Serverless Framework for Multi-Tenant Mixture-of-Experts Serving