换了三台机器，捣鼓了一天时间

环境：
Windows10 / 11
Python 3.10
Paddle 2.5.1 GPU版 （conda install paddlepaddle-gpu==2.5.1 cudatoolkit=11.7 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/ -c conda-forge）
paddlespeech 1.4.1 （pip install paddlespeech）

问题：

1
运行 paddlespeech asr --input zh.wav，先是抛出了一个numpy异常：module 'numpy' has no attribute 'complex'
卸掉1.24，重装1.23解决。

2
继续paddlespeech asr --input zh.wav，另一个问题：
Traceback (most recent call last):
File "D:\Python\envs\Paddle_GPU\lib\site-packages\paddlespeech\cli\asr\infer.py", line 314, in infer
result_transcripts = self.model.decode(
File "D:\Python\envs\Paddle_GPU\lib\site-packages\decorator.py", line 232, in fun
return caller(func, *(extras + args), **kw)
File "D:\Python\envs\Paddle_GPU\lib\site-packages\paddle\fluid\dygraph\base.py", line 347, in _decorate_function
return func(*args, **kwargs)
File "D:\Python\envs\Paddle_GPU\lib\site-packages\paddlespeech\s2t\models\u2\u2.py", line 818, in decode
hyp = self.attention_rescoring(
File "D:\Python\envs\Paddle_GPU\lib\site-packages\paddlespeech\s2t\models\u2\u2.py", line 532, in attention_rescoring
assert speech.shape[0] == speech_lengths.shape[0]
IndexError: list index out of range
KeyError: 'result'

paddlespeech 1.4.1在paddle 2.4.2.post117下是能跑通示例的。但是还有其他乱七八糟依赖库的不兼容问题。