error using pretrained acoustic models (solved, potential confusion in MFA documentation page)

[jiangshizong]

I tried using the pretrained french acoustic model (trained on GP using GP dictionary), but got error messages about not finding the pretrained model.

The MFA documentation page (https://montreal-forced-aligner.readthedocs.io/en/latest/aligning.html#align-using-pretrained-models) says that the pretrained model should be under pretrained_models folder and shouldn't add the .zip suffix as an argument in the command line. However, it seems that this only works for English.

What I did was just placing French.zip directly under Montreal-Forced-Alignment folder and using French.zip as the name of the argument. If you put it under the pretrained_models folder then it won't work.

[YugwonWon]

• After using the mfa g2p command, I used the mfa align command. But in mfa align I get an error.
• I'm getting pretrainedModelNotFoundError on Korean model. The commands I've tried are: You have correctly entered the path of the pretrained model as the third argument. Still it doesn't work.

  align_cmd = ['mfa', 'align', f'{temp_name}/in', f'{temp_name}/dict', '/Users/username/Desktop/projects/mfa/korean_mfa.zip', f'{temp_name }/out', '-t', f'{temp_name}/log', '--clean', '-v', '--config_path', '/Users/yugwon/Desktop/projects/mfa/align_config. yaml', '>', f'{os.path.join(temp_name, "align.log")}']
  os.system(' '.join(align_cmd))

• error msg:

PretrainedModelNotFoundError:

   Could not find a model named "/Users/username/Desktop/projects/mfa/tmp/4376135040/dict" for dictionary.
Traceback (most recent call last):

I am using mac os and I am using miniconda3 installed in python 3.8 environment.