-
tried both grpc_tools_node_protoc --plugin=protoc-gen-ng=.\node_modules\.bin\protoc-gen-ng.cmd --ng_out=src/app/proto -I src/app/proto src/app/proto/geo.proto and protoc --plugin=protoc-gen-ng=.\node_…
-
Hi, I am trying to make training with two dataset, one with 300 h of data and other with 30 h data. One I training with 300 h of data and got pretty good WER, so I attempt to fork this model and train…
-
### Question
I want to add spec augmentation while training to enhance WER
#### Additional Context
I am running experiments on one of the Indian languages. I have followed the tutorials available…
-
### Question
I am getting low thrpt(sec/sec) when training conv_glu with ASG criterion, and one epoch takes 18 hours to run. Could you please let me know if this kind of thing is expected and how to …
-
## 🐛 Bug
It seems there is a device mismatch at these lines, [(one)](https://github.com/pytorch/fairseq/blob/0ffb94151f597ecb677551289e7046a21fb5ebaf/fairseq/models/wav2vec/wav2vec2.py#L480) and [(…
-
I am trying to achieve some good results on Libri 100 Hour data using transformer + CTC architecture provided in https://github.com/facebookresearch/wav2letter/blob/master/recipes/models/sota/2019/li…
-
I've tried to reproduce seq2seq_tds. Following the guide wav2letter/recipes/models/seq2seq_tds/librispeech/READEME.md.
But I got lr=0 at second epoch.
```
I0731 10:19:50.247529 78586 Train.cpp:3…
-
When calling my lambda i have :
`
"errorMessage": "Cannot find module '/var/task/handler'",
"errorType": "Error",
"stackTrace": [
"Function.Module._load (module.js:276:25)"…
-
I train wav2letter on my own language with approximately 100K utterance ( ~10sec/utterance).
I using librespeech tutorial recipes with wav2vec feature instead of MFCC.
It took about 1h30m to complet…
-
Thanks to Facebook for making such a good ASR.
I would like to train librispeech data using tutorial (https://github.com/facebookresearch/wav2letter/tree/master/tutorials/1-librispeech_clean) with C…