Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 20 of 60 for “"mixed precision"”.

  1. Efficient Mixed-Precision Inference for Vision Transformers

    [ES] Los avances recientes en el aprendizaje profundo (DL) se han logrado al ampliar el número de parámetros de los modelos y los volúmenes de datos. El Transformer es la arquitectura más utilizada en el procesamiento del lenguaje natural (NLP), destacando su mecanismo de atención, que permite …

    upv Repository record for Efficient Mixed-Precision Inference for Vision Transformers (opens in a new tab)

  2. Mixed Precision Multi-frame Parallel Low-Density Parity-Check Code Decoder

    … decoder achieves 12.6 Gbps at 3.0 dB SNR. The mixed precision scheme is explored by adding precision control and alignment units before and after check node units (CNU) to improve performance, as well as error floor. By mixing the 6-bit and 5-bit precision CNUs at 1:1 ratio, the decoder reaches …

    tamu Repository record for Mixed Precision Multi-frame Parallel Low-Density Parity-Check Code Decoder (opens in a new tab)

  3. Mixed-Precision Numerical Linear Algebra Algorithms: Integer Arithmetic Based LU Factorization and Iterative Refinement for Hermitian Eigenvalue Problem

    <p>Mixed-precision algorithms are a class of algorithms that uses low precision in part of the algorithm in order to save time and energy with less accurate computation and communication. These algorithms usually utilize iterative refinement processes to improve the approximate solution obtained …

    utk-theses Repository record for Mixed-Precision Numerical Linear Algebra Algorithms: Integer Arithmetic Based LU Factorization and Iterative Refinement for Hermitian Eigenvalue Problem (opens in a new tab)

  4. Energy-Efficient Hardware Acceleration of Transformer-Based Models

    … parameter-wise method assigns optimal bit precision based on sensitivity analysis, while the proposed row-wise quantization enhances hardware efficiency by converting mixed-precision matrices into two uniform-precision blocks. For attention activations, the proposed Weight-Hessian-aware KV …

    gatech Repository record for Energy-Efficient Hardware Acceleration of Transformer-Based Models (opens in a new tab)

  5. High Performance Reconfigurable Computing for Linear Algebra: Design and Performance Analysis

    … significant speedup. Second, computing with mixed-precision data on FPGAs is proposed for higher performance. Experimental analysis shows that mixed-precision algorithms on FPGAs can achieve the high performance of using lower-precision data while keeping higher-precision accuracy for finding …

    utk-theses Repository record for High Performance Reconfigurable Computing for Linear Algebra: Design and Performance Analysis (opens in a new tab)

  6. Using Runtime Floating Point Accuracy Feedback to Make Automated Precision/Performance Improvements or Tradeoffs

    … In addition, we present a profile driven mixed precision analysis framework that heuristically determines the precision of all variables in an application based on their numerical behavior. We experimentally evaluate the mixed precision analysis to show that it can generate a range of …

    duke Repository record for Using Runtime Floating Point Accuracy Feedback to Make Automated Precision/Performance Improvements or Tradeoffs (opens in a new tab)

  7. Optimization of Deep Neural Networks for Efficient Resource Utilization

    … on the Cityscapes dataset. Additionally, mixed-precision QAT was investigated using first-order gradient-based sensitivity analysis to assign per-layer bit-widths. Maintaining activation precision at or above 6 bits during mixed-precision QAT enabled substantial compression—up to …

    chalmers Repository record for Optimization of Deep Neural Networks for Efficient Resource Utilization (opens in a new tab)

  8. AIR: Adaptive Dynamic Precision Iterative Refinement

    … arithmetic units with statically defined precision. This dissertation explores novel approaches to improve computing architectures for linear system applications not only by designing application-specific hardware but also by optimizing precision by applying adaptive dynamic precision …

    utk-theses Repository record for AIR: Adaptive Dynamic Precision Iterative Refinement (opens in a new tab)

  9. Accuracy and Hardware Cost Analysis of Multi-Format Floating-Point Arithmetic Generated by FloPoCo on FPGA

    … a comparative analysis between a unified-precision Nfloat MAC and an IEEE fused multiply-add (FMA) reveals that the NFloat MAC achieves over 50% power and area savings while maintaining identical algorithmic fidelity at low-to-medium precisions. Finally, we investigate the performance of …

    chalmers Repository record for Accuracy and Hardware Cost Analysis of Multi-Format Floating-Point Arithmetic Generated by FloPoCo on FPGA (opens in a new tab)

  10. Accelerated CTIS Using the Cell Processor

    … addition to demonstrating the feasibility of a mixed precision implementation, it is shown that use of the parallel processing capabilities of the Cell may provide a significant reduction in reconstruction time.</p>

    ut-knoxville Repository record for Accelerated CTIS Using the Cell Processor (opens in a new tab)

  11. Precision-cascading in restarted GMRES

    … a multitude of applications. Alongside this, mixed-precision is a growing novel field of methodology which utilizes non-standard precision configurations to modify algorithms and tune their characteristics. In this thesis, we propose a simple inter-iteration approach to applying …

    ubc Repository record for Precision-cascading in restarted GMRES (opens in a new tab) Download PDF for Precision-cascading in restarted GMRES (opens in a new tab)

  12. Hardware Acceleration of Machine Learning

    … Transformer components. We also introduce mixed precision training into our workflow and compare all these optimization techniques step by step with baseline performance. In addition, our analysis includes distributed training with multiple GPUs, and a backpropagation time estimation …

    chalmers Repository record for Hardware Acceleration of Machine Learning (opens in a new tab)

  13. Flexible Multiple-Precision Fused Arithmetic Units for Efficient Deep Learning Computation

    … (GPU) and the computations are done with single-precision floating-point numbers. However, recent works proved that deep learning computation can be accomplished with low precision numbers. The half-precision numbers are becoming more and more popular in deep learning computation due to their …

    sask Repository record for Flexible Multiple-Precision Fused Arithmetic Units for Efficient Deep Learning Computation (opens in a new tab)

  14. Algorithm-hardware co-optimization for neural network efficiency improvement

    … a systolic array structure, to fully support the mixed-precision inference. In BitSystolic, the numerical precision of both weights and activations can be configured in the range of 2b~8b, fulfilling different requirements across mixed-precision models and tasks. Moreover, the design can support …

    duke Repository record for Algorithm-hardware co-optimization for neural network efficiency improvement (opens in a new tab)

  15. Reducing Communication in the Solution of Linear Systems

    … perturbation. Finally, I investigate using mixed precision in GMRES for solving sparse systems, which reduces the volume of data movement, and thus, the pressure on the memory bandwidth.</p>

    utk-theses Repository record for Reducing Communication in the Solution of Linear Systems (opens in a new tab)

  16. Automated Floating-Point Precision Analysis

    … techniques for automated floating-point precision analysis. The contributions include a software framework that enables floating-point program analysis at the binary level, as well as specific techniques for cancellation detection, mixed-precision configuration, and reduced-precision …

    maryland Repository record for Automated Floating-Point Precision Analysis (opens in a new tab)

Page 1 of 3