Mirco Ravanelli

If the problem was adam-related, this has been resolved with pytorch 1.12.1 On Wed, Aug 10, 2022, 4:56 AM Parcollet Titouan ***@***.***> wrote: > This looks like an issue created...

Thank you @ddavidebb , great job! This looks already in a good shape. Beyond the punctual comments done above, let me share a couple of additional suggestions: 1. It looks...

I'm aware of it and I already fixed it locally. I will push the fix today. On Wed, Nov 10, 2021, 3:17 AM Francesco Paissan ***@***.***> wrote: > ***@***.**** commented...

Any idea about how we can fix that? Not sure this is a problem though

Hi @karla3jo, what is the sampling rate of the signals in your dataset?

How big is your dataset and how many speakers do you have?  If the dataset is not very big, my suggestion is to pretrain the embedding model with the one...

Hi Andrea, maybe we can take the profiling test out and run them only periodically (just like what we do with other tests like the recipe tests, etc). What do...

I'm able to replicate the error as well when I install pytorch 1.9 (and torchaudio 0.9). Any idea Andreas? On Fri, Jul 29, 2022 at 10:53 AM fosimoes ***@***.***> wrote:...

@pplantinga

Hi @csukuangfj, thank you very much for the detailed analysis on the decoding time. We can eventually add this information in the readme file. Right now, `asr-transformer-transformerlm-librispeech` is the best...