multilingual_speech_recognition ValueError: token_ids [] load error.

[monkeycc]

paddlepaddle-gpu 3.1.0 paddlex 3.2.0 Python 3.11.13

from paddlex import create_model
model = create_model(model_name="whisper_large",model_dir=r"I:/AI/PaddleX/yin/whisper_large")
output = model.predict(input=r"I:\AI\PaddleX\yin\zh.wav", batch_size=1)
for res in output:
    res.print()

Adding <|startoflm|> to the vocabulary
Adding <|startofprev|> to the vocabulary
Adding <|nospeech|> to the vocabulary
Adding <|notimestamps|> to the vocabulary
W0831 11:46:32.789902 24164 gpu_resources.cc:243] WARNING: device: 0. The installed Paddle is compiled with CUDNN 9.9, but CUDNN version in your machine is 9.7, which may cause serious incompatible bug. Please recompile or reinstall Paddle with compatible CUDNN version.
Detected language: Chinese
[00:00.580 --> 00:02.580] 我认为跑步最重要的就是
Traceback (most recent call last):
  File "i:\AI\PaddleX\yin\1.py", line 139, in <module>
    for res in output:
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\model.py", line 61, in predict
    yield from self._predictor(*args, **kwargs)
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\base\predictor\base_predictor.py", line 219, in __call__
    yield from self.apply(input, **kwargs)
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\base\predictor\base_predictor.py", line 277, in apply
    prediction = self.process(batch_data, **kwargs)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\multilingual_speech_recognition\predictor.py", line 117, in process
    result = self.model.transcribe(
             ^^^^^^^^^^^^^^^^^^^^^^
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\utils\deps.py", line 139, in _wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\multilingual_speech_recognition\processors.py", line 1024, in transcribe
    add_segment(
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\multilingual_speech_recognition\processors.py", line 951, in add_segment
    text = tokenizer.decode(
           ^^^^^^^^^^^^^^^^^
  File "D:\anaconda3\envs\paddle_3.11\Lib\site-packages\paddlex\inference\models\multilingual_speech_recognition\processors.py", line 243, in decode
    raise ValueError(f"token_ids {token_ids} load error.")
ValueError: token_ids [] load error.

[zxcd]

Which Windows system is this？

[monkeycc]

'CUDA版本': '12.9', 'CPU型号': 'Intel Core Ultra 9 285K', 'Windows': 'Microsoft Windows 11 Professional (x64), Version 24H2, Build 26100.4351', '显卡': '5090D',

[zxcd]

建议尝试paddlex=3.1.0或者develop版本

[monkeycc]

20250911 develop版本 也是一样 @zxcd