Skip to main content

fasr-paraformer

基于 Paraformer 的语音识别流水线插件,为 fasr 提供离线/流式 ASR、VAD 和标点恢复能力。

安装

# 作为 fasr 工作空间成员安装
uv sync

# 或单独安装
pip install fasr-paraformer

快速开始

import fasr

# 加载预置流水线(自动下载模型)
asr = fasr.load("paraformer")

# 识别音频
audios = asr("example.wav")
print(audios[0].text)

预置流水线

流水线名称 组件构成 说明
paraformer FSMN VAD → Paraformer → CT-Transformer 通用中文离线识别,带时间戳
seaco_paraformer FSMN VAD → Seaco-Paraformer → CT-Transformer 支持热词的离线识别
sensevoice FSMN VAD → SenseVoice → CT-Transformer 多语言情感识别
# 带热词的识别
asr = fasr.load("seaco_paraformer")
audios = asr("meeting.wav", hotwords=["Paraformer", "语音识别"])

模型组件

ASR 模型

注册名 默认 checkpoint 说明
paraformer ParaformerForASR iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch 离线 Paraformer,输出带时间戳的 token
seaco_paraformer SeacoParaformerForASR iic/speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch 支持热词偏置的 Paraformer
stream_paraformer ParaformerForStreamASR iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online 流式 Paraformer (Torch)
stream_paraformer.onnx ParaformerForStreamASROnnx 同上 流式 Paraformer (ONNX)

VAD 模型

注册名 说明
fsmn FSMNForVAD FSMN 离线 VAD,ONNX Runtime 推理,内置模型权重
stream_fsmn FSMNForStreamVAD FSMN 流式 VAD (Torch)
stream_fsmn.onnx FSMNForStreamVADOnnx FSMN 流式 VAD (ONNX)

标点模型

注册名 默认 checkpoint 说明
ct_transformer CTTransformerForPunc iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch CT-Transformer 标点恢复,中英文混合

自定义流水线

from fasr import AudioPipeline

pipeline = (
    AudioPipeline()
    .add_pipe("detector", model="fsmn")
    .add_pipe("recognizer", model="seaco_paraformer")
    .add_pipe("sentencizer", model="ct_transformer")
)

audios = pipeline("audio.wav")

依赖

  • fasr >= 0.3.7
  • funasr
  • funasr-onnx >= 0.4.1
  • onnxruntime >= 1.24.3
  • torch >= 2.8.0
  • torchaudio >= 2.8.0
  • Python 3.10–3.12

项目结构

src/fasr_paraformer/
├── __init__.py              # 流水线构建函数 & entry points
├── asset/fsmn-vad/          # 内置 FSMN VAD 模型权重与配置
├── models/
│   ├── asr.py               # Paraformer / Seaco-Paraformer
│   ├── vad.py               # FSMN VAD
│   ├── pun.py               # CT-Transformer 标点恢复
│   ├── stream_asr.py        # 流式 Paraformer (Torch / ONNX)
│   └── stream_vad.py        # 流式 FSMN VAD (Torch / ONNX)
├── preprocessors/
│   ├── waveform/             # Fbank 特征提取、LFR、CMVN
│   └── text/                 # SentencePiece / Char 分词器
└── runtimes/
    └── ort.py                # ONNX Runtime 推理封装

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

fasr_paraformer-0.3.7.post0.tar.gz (3.2 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

fasr_paraformer-0.3.7.post0-py3-none-any.whl (3.2 MB view details)

Uploaded Python 3

File details

Details for the file fasr_paraformer-0.3.7.post0.tar.gz.

File metadata

  • Download URL: fasr_paraformer-0.3.7.post0.tar.gz
  • Upload date:
  • Size: 3.2 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.17 {"installer":{"name":"uv","version":"0.9.17","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for fasr_paraformer-0.3.7.post0.tar.gz
Algorithm Hash digest
SHA256 722ad8643c8cf913af9235b1bae2309095fbe002da82a0cb19cf8cc60e0eda3c
MD5 10456b59ffee4119609110426b1d0c8a
BLAKE2b-256 796e907755c84e731d7f44ee317fee522ba8b1c158fbc29092f3fa97d4f725b6

See more details on using hashes here.

File details

Details for the file fasr_paraformer-0.3.7.post0-py3-none-any.whl.

File metadata

  • Download URL: fasr_paraformer-0.3.7.post0-py3-none-any.whl
  • Upload date:
  • Size: 3.2 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.17 {"installer":{"name":"uv","version":"0.9.17","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for fasr_paraformer-0.3.7.post0-py3-none-any.whl
Algorithm Hash digest
SHA256 64b0e9ee4e990bc5ce121def4bf76a719e2791a44e959f0655cfe14ae671d9c5
MD5 bf5659e79f9cf7076a04df78be26e10b
BLAKE2b-256 12c62322d59b14e340abe88d5a5398166da598f1e6dd9774c389021318c757dd

See more details on using hashes here.

Release history Release notifications | RSS feed

This release

0.3.7.post0 This release

2 files

0.3.7

2 files

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page