UniAttn: Reducing Inference Costs via Softmax Unification for Post-Training LLMs

Documenti analoghi

• Temporal Scaling Law for Large Language Models
  di: Xiong, Yizhe, et al.
  Pubblicazione: (2024)
• Scaffold-BPE: Enhancing Byte Pair Encoding for Large Language Models with Simple and Effective Scaffold Token Removal
  di: Lian, Haoran, et al.
  Pubblicazione: (2024)
• PYRA: Parallel Yielding Re-Activation for Training-Inference Efficient Task Adaptation
  di: Xiong, Yizhe, et al.
  Pubblicazione: (2024)
• Finedeep: Mitigating Sparse Activation in Dense LLMs via Multi-Layer Fine-Grained Experts
  di: Pan, Leiyu, et al.
  Pubblicazione: (2025)
• DSMoE: Matrix-Partitioned Experts with Dynamic Routing for Computation-Efficient Dense LLMs
  di: Lv, Minxuan, et al.
  Pubblicazione: (2025)