jiwer with wrong numpy version induces hang without reported wrong information

[L-Absente]

Firstly, with jiwer=1.3.2 and numpy=2.0.2,
wer = jiwer.wer(truth, hypothesis, standardize=True) would stuck in
Traceback (most recent call last):
File "/root/conformer/EfficientConformer/main.py", line 227, in
main(args)
File "/root/conformer/EfficientConformer/main.py", line 128, in main
model.fit(dataset_train,
^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/conformer/EfficientConformer/models/model.py", line 313, in fit
wer, truths, preds, val_loss = self.evaluate(dataset, val_steps, verbose_val, eval_loss=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/conformer/EfficientConformer/models/model.py", line 502, in evaluate
wer = jiwer.wer(speech_true, speech_pred, standardize=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/lib/python3.12/site-packages/jiwer/wer.py", line 75, in wer
distance = _edit_distance(t, h)
^^^^^^^^^^^^^^^^^^^^
File "/root/miniconda3/lib/python3.12/site-packages/jiwer/wer.py", line 252, in _edit_distance
m[i, j] = min(
^^^^
But it never reported wrong information

Solution:
update jiwer's version and
change
wer = jiwer.wer(truth, hypothesis, standardize=True)
into
transformation = jiwer.Compose([
jiwer.RemoveWhiteSpace(),
jiwer.ToLowerCase(),
jiwer.ReduceToListOfListOfWords(word_delimiter=" ")
])
wer = jiwer.wer(truth, hypothesis, truth_transform=transformation, hypothesis_transform=transformation)