Ladder-residual: parallelism-aware architecture for accelerating large model inference with communication overlapping

Documenti analoghi

• FlashFormer: Whole-Model Kernels for Efficient Low-Batch Inference
  di: Nrusimha, Aniruddha, et al.
  Pubblicazione: (2025)
• OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization
  di: Zhou, Zhongzhu, et al.
  Pubblicazione: (2026)
• Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time
  di: Zhang, Zhenyu, et al.
  Pubblicazione: (2025)
• Reducing Transformer Key-Value Cache Size with Cross-Layer Attention
  di: Brandon, William, et al.
  Pubblicazione: (2024)
• Improving Model Alignment Through Collective Intelligence of Open-Source LLMS
  di: Wang, Junlin, et al.
  Pubblicazione: (2025)