Published March 13, 2024 | Version v1
Model Open

English Espnet speech recognition model trained on librispeech

Authors/Creators

Contributors

Description

This is the baseline tranformer model for English Speech Recognition trained using ESPNET v1 on the LibriSpeech database. 

Performances are as below on the dev_clean subset:

write a CER (or TER) result in exp/train_960_lc.rm_pytorch_train_transformer_large_unigram5000_specaug/decode/test/exp/train_960_pytorch_train_transformer_large_specaug/results/model.val5.avg
.best/result.txt
|       SPKR               |       # Snt              # Wrd        |       Corr                 Sub                Del                 Ins                 Err               S.Err        |
|       Sum/Avg            |       2703               67572        |       96.2                 2.9                0.9                 0.6                 4.4                37.5        |
write a WER result in exp/train_960_lc.rm_pytorch_train_transformer_large_unigram5000_specaug/decode/test/exp/train_960_pytorch_train_transformer_large_specaug/results/model.val5.avg.best/res
ult.wrd.txt
|       SPKR               |        # Snt              # Wrd        |       Corr                  Sub                 Del                 Ins                  Err               S.Err        |
|       Sum/Avg            |        2703               54402        |       96.8                  2.9                 0.3                 0.5                  3.6                37.4        |

Files

baseline_librispeech.large.espnet1.zip

Files (361.3 MB)

Name Size Download all
md5:55297b6c7bb22778b2cd809d53613575
361.3 MB Preview Download

Additional details

Funding

European Commission
FVLLMONTI - Ferroelectric Vertical Low energy Low latency low volume Modules fOr Neural network Transformers In 3D 101016776

References

  • L. Ben Letaifa and J.-L. Rouas, 'Transformer Model Compression for End-to-End Speech Recognition on Mobile Devices', in EUSIPCO 2022, 2022.