sherpa
1.3
  • Introduction
  • Download pdf
  • Social groups
  • Run Next-gen Kaldi in your browser
  • Pre-trained models

k2-fsa/sherpa

  • sherpa

k2-fsa/sherpa-ncnn

  • sherpa-ncnn

k2-fsa/sherpa-onnx

  • sherpa-onnx
    • Tutorials
    • Installation
    • Frequently Asked Question (FAQs)
    • Python
    • C API
    • Java API
    • Javascript API
    • Kotlin API
    • Swift API
    • Go API
    • C# API
    • Pascal API
    • Lazarus
    • WebAssembly
    • Android
    • HarmonyOS
    • iOS
    • Flutter
    • WebSocket
    • Hotwords (Contextual biasing)
    • Keyword spotting
    • Punctuation
    • Audio tagging
    • Spoken language identification
    • VAD
    • Pre-trained models
    • Moonshine
    • SenseVoice
    • Paraformer
    • NeMo
    • FireRedAsr
    • Dolphin
    • 拼音词组匹配替换
    • Speaker Diarization
    • Speaker Identification
    • Speech enhancement
    • Source separation
    • rknn
    • Ascend NPU (昇腾 NPU)
      • Install
      • Export models to Ascend NPU
      • Pre-trained models
    • Text-to-speech (TTS)

Triton

  • Triton
sherpa
  • »
  • sherpa-onnx »
  • Ascend NPU (昇腾 NPU)
  • Edit on GitHub

Ascend NPU (昇腾 NPU)

This section describes how to use sherpa-onnx with Ascend NPU from Huawei.

The following NPUs are known to work:

  • 910B

  • 310P3

Other types of Ascend NPU may also work.

  • Install
    • Install CANN
    • Build sherpa-onnx
  • Export models to Ascend NPU
    • Export SenseVoice
    • Export Paraformer
  • Pre-trained models
    • sherpa-onnx-ascend-910B-sense-voice-zh-en-ja-ko-yue-2024-07-17 (Chinese, English, Japanese, Korean, Cantonese, 中英日韩粤语)
      • Decode long files with a VAD
      • Decode a short file
    • sherpa-onnx-ascend-910B-sense-voice-zh-en-ja-ko-yue-2025-09-09 (Chinese, English, Japanese, Korean, Cantonese, 中英日韩粤语)
Previous Next

© Copyright 2022-2025, sherpa development team.

Built with Sphinx using a theme provided by Read the Docs.