Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
FreeScale: Distributed Training for Sequence Recommendation Models with Minimal Scaling Cost
Chenhao Feng, Haoli Zhang, Shakhzod Ali-Zade +17
Modern industrial Deep Learning Recommendation Models typically extract user preferences through the analysis of sequential interaction histories, subsequently generating predictio…
cs.LG2026
MoEBlaze: Breaking the Memory Wall for Efficient MoE Training on Modern GPUs
Jiyuan Zhang, Yining Liu, Siqi Yan +6
The pervasive "memory wall" bottleneck is significantly amplified in modern large-scale Mixture-of-Experts (MoE) architectures. MoE's inherent architectural sparsity leads to spars…