使用owsm_v4_medium_1B进行ASR任务

[newday1255]

环境：
torch 2.7.1+cu126
espnet-model-zoo 0.1.7
espnet 202509
librosa 0.11.0

import soundfile
from espnet2.bin.asr_inference import Speech2Text
speech2text = Speech2Text.from_pretrained(
"/data/espnet/owsm_v4_medium_1B",
device="cuda",
beam_size=5,
ctc_weight=0.0,
maxlenratio=1.0,
lang_sym="",
task_sym="",
predict_time=False,
)
speech, rate = soundfile.read("audio.wav")
nbests = speech2text(speech)

text, *_ = nbests[0]
print(text)

这里已经把owsm_v4_medium_1B模型文件从huggingface上下载到了本地，下载的模型里面有文件夹data、exp，其中模型文件是valid.total_count.ave_5best.pth，
报错的信息如下：
File "/data/projects/owsm/espnet/owsm_test.py", line 3, in
speech2text = Speech2Text.from_pretrained(
File "/data/projects/owsm/espnet/espnet2/bin/asr_inference.py", line 709, in from_pretrained
kwargs.update(**d.download_and_unpack(model_tag))
File "/opt/miniconda/lib/python3.9/site-packages/espnet_model_zoo/downloader.py", line 372, in download_and_unpack
return self.unpack_local_file(url)
File "/opt/miniconda/lib/python3.9/site-packages/espnet_model_zoo/downloader.py", line 244, in unpack_local_file
return unpack(filename, outdir)
File "/data/projects/owsm/espnet/espnet2/main_funcs/pack_funcs.py", line 199, in unpack
with Archiver(input_archive) as archive:
File "/data/projects/owsm/espnet/espnet2/main_funcs/pack_funcs.py", line 32, in init
raise ValueError(f"Cannot detect archive format: type={file}")
ValueError: Cannot detect archive format: type=/opt/miniconda/lib/python3.9/site-packages/espnet_model_zoo/2bd56a8590f881bbf2b93f34fe1421dd/owsm_v4_medium_1B

请教如何解决这个问题