Practical FP4 Training for Large-Scale MoE Models on Hopper GPUs

Documenti analoghi

• FP8-Flow-MoE: A Casting-Free FP8 Recipe without Double Quantization Error
  di: Wang, Fengjuan, et al.
  Pubblicazione: (2025)
• Grouter: Decoupling Routing from Representation for Accelerated MoE Training
  di: Xu, Yuqi, et al.
  Pubblicazione: (2026)
• MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs
  di: Cao, Shiyi, et al.
  Pubblicazione: (2024)
• MoEBlaze: Breaking the Memory Wall for Efficient MoE Training on Modern GPUs
  di: Zhang, Jiyuan, et al.
  Pubblicazione: (2026)
• LocMoE: A Low-Overhead MoE for Large Language Model Training
  di: Li, Jing, et al.
  Pubblicazione: (2024)