Error: can't load model

[janson91]

hi, I tried some wav2vec models, but there was always error:
the code is
`import torch
import fairseq

cp_path = '/path/to/wav2vec.pt'
model, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([cp_path])
model = model[0]
model.eval()`

and the error :
the model: https://dl.fbaipublicfiles.com/fairseq/wav2vec/wav2vec_large.pt Error:KeyError: 'speech_pretraining'
the model:https://dl.fbaipublicfiles.com/fairseq/wav2vec/vq-wav2vec.pt Error:KeyError: 'binary_cross_entropy'
the model:https://dl.fbaipublicfiles.com/fairseq/wav2vec/vq-wav2vec_kmeans.pt KeyError: 'binary_cross_entropy'
the model:https://dl.fbaipublicfiles.com/fairseq/wav2vec/xlsr_53_56k.pt AttributeError: 'NoneType' object has no attribute 'task'

is anywhere more examples for loading the model?

thanks

[myleott]

What version of fairseq do you have installed? You may need to install the latest version from source: https://github.com/pytorch/fairseq/#requirements-and-installation

[olafthiele]

@janson91 Check the colab mentioned in this issue. You might have to checkout that days commit, the newer ones seem to have problems. An reload the xlsr..pt if it is older than 7 days. There was a problem with it.

[yc-li20]

same issue though fairseq is the latest version.

[duwf2003]

I have solved this problem.
(1)You need to find out where is the location of fairseq, in my env is ~/miniconda3/envs/myconda/lib/python3.7/site-packages/fairseq/.
(2)After that, you need to change your location, and modify the file checkpoint_utils.py
cd ~/miniconda3/envs/myconda/lib/python3.7/site-packages/fairseq/
vi checkpoint_utils.py
line 438
if choice is not None:
change to
if choice is not None and choice in REGISTRY["registry"]:

(3) After save the checkpoint_utils.py, this is the workable example:
cp = 'path/to/vq-wav2vec.pt'
model, cfg, task = fairseq.checkpoint_utils.load_model_ensemble_and_task([cp])
model = model[0]
model.eval()

[stale]

This issue has been automatically marked as stale. If this issue is still affecting you, please leave any comment (for example, "bump"), and we'll keep it open. We are sorry that we haven't been able to prioritize it yet. If you have any new additional information, please include it with your comment!

[stale]

Closing this issue after a prolonged period of inactivity. If this issue is still present in the latest release, please create a new issue with up-to-date information. Thank you!