How to use OWSM model? Getting TypeError in Speech2Text.from_pretrained()

[cifkao]

I'm excited to try the OWSM 3.1 model, but the snippet provided on HuggingFace under "Use in ESPnet" does not work for me:

from espnet2.bin.asr_inference import Speech2Text

model = Speech2Text.from_pretrained(
  "espnet/owsm_v3.1_ebf"
)

I get the following error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[1], line 3
      1 from espnet2.bin.asr_inference import Speech2Text
----> 3 model = Speech2Text.from_pretrained(
      4   "espnet/owsm_v3.1_ebf"
      5 )

File ~/mambaforge/envs/espnet/lib/python3.10/site-packages/espnet2/bin/asr_inference.py:679, in Speech2Text.from_pretrained(model_tag, **kwargs)
    676     d = ModelDownloader()
    677     kwargs.update(**d.download_and_unpack(model_tag))
--> 679 return Speech2Text(**kwargs)

TypeError: Speech2Text.__init__() got an unexpected keyword argument 's2t_train_config'

I tried espnet/owsm_v3 too and it doesn't work either.

What is the correct code to load the model and run transcription? Thanks!

[sw005320]

It should be espnet2.bin.s2t_inference not espnet2.bin.asr_inference

@pyf98, we encountered this misuse in two cases.

• Is there a way to avoid this issue? (at least provide some more meaningful error messages?)
• We have to update "Use in ESPnet" in

This actually shows `espnet2.bin.asr_inference`

[pyf98]

yeah, I see. We need to update "Use in ESPnet". I think it is defined in Hugging Face not ESPnet?