Mirco Ravanelli

@jamfly, do you have some update on this PR (e.g, other recipes)?

Thanks for this PR. Results look pretty good. Let me review it as soon as I can.

Right, not sure. I forgot about that but, yes, NAR is something important that we might want to consider for other recipes (e.g, LibriSpeech) On Wed, 10 Nov 2021 at...

Makes sense! On Wed, 10 Nov 2021 at 14:16, Parcollet Titouan ***@***.***> wrote: > I'll review the code of this PR, but unsure about running the experiment > ... We...

That happened to me with some other models. Unfortunately, this might happen because when you change seed you also change the initialization point and some initializations might be more problematic...

Hi @kunnark, thank you for this contribution. What is the performance you got with this model?

Hi @kunnark, I just push some minor style issues. I would like to try this recipe as soon as they resume my cluster. As you might know, we normally share...

Hi @kunnark, I'm testing this PR. I have the following comments: 1. It looks like there is a problem with the wav2vect part: ``` Traceback (most recent call last): File...

@ycemsubakan

Hi, thank you for sharing your results. I'm not sure I totally understood your problem. However, if you use a pre-trained model (e.g., on English) and you fine-tune it on...