-
**Describe the bug**
I am trying to finetune the XXL Conformer-Transducer model from a base .nemo checkpoint that I downloaded from https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/stt_…
-
Is it ok that default model (I use ASR conformer transducer) has WER=0.26 but the same model that I have been training for 2 days has WER=0.32.
I use [Russian version](https://huggingface.co/nvidia/s…
-
微博内容精选
-
**Describe the bug**
There is a size mismatch between EncDec and Joint and Decoder modules when trying to finetune a Large Conformer-Transducer model.
**Steps/Code to reproduce bug**
I instal…
-
**Describe the bug**
I am training a conformer-transducer medium model and on the validation set I reach `val_wer=0.0157-epoch=13.ckpt`, where val_wer is actualy ChER as i use `use_cer=True`. When …
-
Hi,
I am new to Espnet and looking for streaming ASR support either with RNN-T or Transformer transducer.
Can someone guide me to the appropriate recipe to start for this and the documentation if …
-
**Describe the bug**
Thanks to you guys I have succesfully finetuned a fast conformer transducer model and it is performing quite well. I wanted to see if I am able to continue finetuning for the C…
-
I downloaded pretrained librispeech models from https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/tree/main and using them for decoding through p…
-
Hello,
We are currently using a Bengali ASR model which we trained using espnet's Librispeech recipe [(config).](https://github.com/espnet/espnet/blob/master/egs/librispeech/asr1/conf/tuning/train…
-
Hi,
I am using the Voxforge based experiment with RNN-T and I've set language to english in run_rnnt.sh
The training completes and during decoding I am receiving the following errors:
```
Error: N…