Skip to content
Discussion options

You must be logged in to vote

结论:在当前 FunASR v1.4.3 / main 中,AutoModelVLLM 本身还不能通过 vad_modelspk_modelsentence_timestamp=True 组装 sentence_info。你现在只拿到 token/字级 timestamps 是当前实现的预期结果,不是这些参数写错了。

原因是 AutoModelVLLM 当前直接包装 FunASRNanoVLLM:后者的 generate() 只在 CTC 可用时添加 timestamps,没有运行 VAD、CAM++ 或 sentence_info 后处理;上述额外参数被 **kwargs 接收,但不会启用这条组合 pipeline。

如果当前业务必须使用稳定的句级时间戳/说话人输出,请先使用公开的 AutoModel pipeline:

from funasr import AutoModel

model = AutoModel(
    model="FunAudioLLM/Fun-ASR-Nano-2512",
    vad_model="fsmn-vad",
    punc_model="ct-punc",
    spk_model="cam++",
    device="cuda",
)
result = model.generate(
    input="audio.wav",
    sentence_timestamp=True,
    return_spk_res=True,
)
print(result[0]["sentence_info"])

源…

Replies: 1 comment 1 reply

Comment options

You must be logged in to vote
1 reply
@Stephanie-ting
Comment options

Answer selected by LauraGPT
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
None yet
2 participants