Import stt is not working! about self-supervised-speech-recognition HOT 7 OPEN

Are you using the right fairseq git checkpoint?

git clone https://github.com/pytorch/fairseq.git
cd fairseq
git checkout c8a0659be5cdc15caa102d5bbf72b872567c4859
pip install --editable ./
cd ..

from self-supervised-speech-recognition.

Yes

from self-supervised-speech-recognition.

You can look at some threads here:
https://stackoverflow.com/questions/58585690/python-cannot-import-unknown-location
https://stackoverflow.com/questions/63728242/importerror-cannot-import-name-unknown-location

from self-supervised-speech-recognition.

from self-supervised-speech-recognition.

Make sure the followings are correct:

• you install the fairseq correctly, which means you can import the lib from any directory.
• the fairseq dir is not located in the same directory with stt.py
• You calling the "from stt import Transcriber" under "self-supervised-speech-recognition" directory.

from self-supervised-speech-recognition.

hi,

I have installed Fairseq successfully!
2) I can't import Fairseq from any directory but only when i am inside fairseq directory
3) I am calling "from stt import Transcriber" under "self-supervised-speech-recognition" directory but i am getting the error """" ImportError: cannot import name 'Transcriber' from 'stt' (/content/self-supervised-speech-recognition/stt.py)""""
4) Note that i have to make changes to the stt.py files which are as follows:

from fairseq.fairseq import checkpoint_utils, options, progress_bar, tasks, utils
from fairseq.fairseq.data.data_utils import post_process
from fairseq.fairseq.logging.meters import StopwatchMeter, TimeMeter

This is because when i am following your instructions, the folder structure is fairseq>fairseq>checkpoint_utils and so on.

Could you please check my script:

https://github.com/gauravgund/gauravgund/blob/master/wav2vec_fairseq__finetuning.ipynb

from self-supervised-speech-recognition.

@mailong25 @gauravgund got the same issue, but it works fine after restarted. We can close it now.

from self-supervised-speech-recognition.