search

nii-yamagishilab / mos-finetune-ssl

BSD 3-Clause "New" or "Revised" License
82 stars 20 forks source link

Loading pre-trained wav2vec2.0 model error #1

Closed Kristopher-Chen closed 2 years ago

Kristopher-Chen commented 2 years ago

Hi, have you encountered the error of loading pre-trained wav2vec2.0 models?

the code is as follows: cp_path = './ssl/xlsr_53_56k.pt' model, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([cp_path]) model = model[0] model.eval()

it works with hubert_large_ll60k.pt, but not with xlsr_53_56k.pt.

the error code is omegaconf.errors.ConfigKeyError: Key 'eval_wer' not in 'AudioPretrainingConfig' full_key: eval_wer reference_type=Optional[AudioPretrainingConfig] object_type=AudioPretrainingConfig

Thank you!

liushenme commented 2 years ago

Hi, I have the same problem. Have you solved it?

Kristopher-Chen commented 2 years ago

Hi, I have the same problem. Have you solved it?

Yes, there is something missing in the defined structure. Copy the missing definitions from fairseq/tasks/audio_fintuning.py to fairseq/tasks/audio_pretraining.py