DyBit: Dynamic Bit-Precision Numbers for Efficient Quantized Neural Network Inference

Ähnliche Einträge

• A Composable Dynamic Sparse Dataflow Architecture for Efficient Event-based Vision Processing on FPGA
  von: Gao, Yizhao, et al.
  Veröffentlicht: (2024)
• TATAA: Programmable Mixed-Precision Transformer Acceleration with a Transformable Arithmetic Architecture
  von: Wu, Jiajun, et al.
  Veröffentlicht: (2024)
• Co-designing a Sub-millisecond Latency Event-based Eye Tracking System with Submanifold Sparse CNN
  von: Zhang, Baoheng, et al.
  Veröffentlicht: (2024)
• Quantization Variation: A New Perspective on Training Transformers with Low-Bit Precision
  von: Huang, Xijie, et al.
  Veröffentlicht: (2023)
• Beyond Outliers: A Data-Free Layer-wise Mixed-Precision Quantization Approach Driven by Numerical and Structural Dual-Sensitivity
  von: Zhang, Hengyuan, et al.
  Veröffentlicht: (2026)