ggml-org/whisper.cpp

NPU support in whisper.cpp

Open

#1 557 ouverte le 27 nov. 2023

Voir sur GitHub
 (18 commentaires) (12 réactions) (0 assignés)C++ (49 693 stars) (5 535 forks)batch import
good first issueperformanceresearch🔬

Description

Christmas is coming soon, and I want to take some time to research something interesting, such as edge low-power inference. Although current whisper.cpp can run on Raspberry Pi, the inference performance cannot achieve real-time transcription. Fortunately, there are now some development boards that use processors with NPUs, which can be used to achieve real-time transcription of large models. My primary goal is to first support RK3566 and RK3588.

Roadmap:

  • MatMul offloading
  • Conv-Gelu offloading
  • LayerNorm offloading ...

Reference:

https://github.com/rockchip-linux/rknpu2

Guide contributeur

Stack technique
cpp
Domaine
machine learningperformance
Type d'issue
feature
DifficultéDifficulté estimée pour un nouveau contributeur, de 1 pour un très petit changement à 5 pour un travail expert.
5
Temps estiméFourchette de temps approximative pour investiguer, implémenter, tester et préparer une pull request.
over 1 week
Statut d'activitéDisponibilité apparente de l'issue : fraîche, active, ancienne, bloquée ou en attente d'un mainteneur.
active
ClartéClarté avec laquelle l'issue explique le changement attendu, les critères d'acceptation et la prochaine étape.
mostly clear
Prérequis
C++ programmingNeural network inference basicsNPU architecture knowledgeRK3566/RK3588 familiarity
Accessibilité débutantScore de 1 à 100 estimant l'accessibilité de cette issue pour un premier contributeur.
20
Direction de recherche
The issue proposes adding NPU support for RK3566 and RK3588, with a roadmap for offloading MatMul, Conv Gelu, and LayerNorm. Research should focus on integrating with the Rockchip RKNN toolkit (rknpu2) and modifying the whisper.cpp backend to delegate these operations to the NPU. Investigate existing ggml backend implementations for similar hardware and adapt the build system. Review the linked discussion on Raspberry Pi for performance baselines.