2 papers
cs.DC2026
Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs
Haoyu Zhang, Arshad Rahman, Adnan Aziz +55
Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history o…
cs.IR2024
Disaggregating Embedding Recommendation Systems with FlexEMR
Yibo Huang, Zhenning Yang, Jiarong Xing +5
Efficiently serving embedding-based recommendation (EMR) models remains a significant challenge due to their increasingly large memory requirements. Today's practice splits the mod…