Wav2letter: Current pre-trained model?

[nh2]

The current version seems to no longer link the pre-trained model mentioned in #29 #88.

Is there a new/current pre-trained model available?

[vineelpratap]

Hi,
Currently it is not available. We plan to open-source within few weeks.

[nh2]

Thanks, eagerly looking forward to it!

When it's done, could you post here, so that subscribers will notice?

[zdgithub]

As there isn't a pre-trained model,
how many epochs do I need to achieve the result on librispeech dataset in paper?

[pastorenick]

Hey guys, is there a reason why pre-trained models are not accessible anymore anywhere?

Hi,
Currently it is not available. We plan to open-source within few weeks.

This was posted on the 10th of January, BTW.

[Venryx]

Any update?

If a pre-trained model was close to being available all the way back in January, surely some progress has been made on that front by this point?

[tlikhomanenko]

@Venryx

Right now, please, check them here https://github.com/facebookresearch/wav2letter/tree/master/models

[gemfield]

@tlikhomanenko this link is broken too, any update?

[tlikhomanenko]

Hi @gemfield,

We restructured a little bit folders. So now you can find everything about reproduce of papers here https://github.com/facebookresearch/wav2letter/tree/master/recipes/models, for example released models for seq2seq_tds paper for librispeech here https://github.com/facebookresearch/wav2letter/tree/master/recipes/models/seq2seq_tds/librispeech. For lexicon-free you can find models here https://github.com/facebookresearch/wav2letter/tree/master/recipes/models/lexicon_free (these acoustic models are similar to ones used in conv_glu paper)