AssertionError: paraformer-zh-streaming is not registered

[bjxx-algo]

🐛 Bug

To Reproduce

[Steps to reproduce the behavior. Always include the exact command you ran.

Name Version Build Channel

funasr 1.2.9 pypi_0 pypi
modelscope 1.33.0 pypi_0 pypi

Code sample

https://huggingface.co/funasr/paraformer-zh-streaming

follow the script in the model card

Expected behavior

Error logs

funasr/auto/auto_model.py", lin

e 125, in init
model, kwargs = self.build_model(**kwargs)
File "/data1051n/algorithm/home/luor/modelzoos/FunASR/funasr/auto/auto_model.py", line 264, in build_model
assert model_class is not None, f'{kwargs["model"]} is not registered'
AssertionError: funasr/paraformer-zh-streaming is not registered

Environment

• OS:
• Python version:
• FunASR version:
• ModelScope version:
• PyTorch / torchaudio version: 2.5
• Install method (pip, source, Docker): pip
• Device (cuda, cpu, mps):
• GPU model:
• CUDA/cuDNN version:
• Docker image tag, if used:

Audio details

If the audio cannot be shared, please describe:

• Duration:
• Sample rate:
• Format:
• Language/dialect:
• Speaker count:
• Background noise/music:

[LauraGPT]

This issue is caused by using an outdated FunASR version (1.2.9). The model funasr/paraformer-zh-streaming loads correctly from HuggingFace in the latest version.

Verified on FunASR 1.3.7:

from funasr import AutoModel

model = AutoModel(model="funasr/paraformer-zh-streaming", hub="hf", device="cuda")
res = model.generate(input="audio.wav")
# Output: "欢迎大家来体验达摩院推出的语音识别模型..."

Fix: Please upgrade FunASR:

pip install -U funasr

The current version (1.3.7) fully supports loading models from HuggingFace via the hub="hf" parameter.