Hervé BREDIN comments

Results 270 comments of


                                            Hervé BREDIN

add pytorch example

I checked: using `DataLoader` like I described above, or using `pescador.maps.buffer_stream` is more or less equivalent in terms of performance. For some reason, I thought `DataLoader` would be smarter than...

feat: add SpectraclClustering back

cc @wq2012 -- I plan to bring back spectral clustering (with a limited set of hyper-parameters) in version 2.1

Pipeline inference with gpu?

It does run on GPU by default when GPUs are available.

Pipeline inference with gpu?

Yes.

TPU inference?

I would like to know that too! What did you try?

TPU inference?

To check if inference is supported, you could start simple and run something like this and see if it complains about TPUs. ```python from pyannote.audio import Inference inference_on_gpu = Inference("pyannote/segmentation",...

TPU inference?

Without an actual access to TPUs, I am afraid I won't be able to help you here....

TPU inference?

Sure! The last time I tried getting access to a TPU on Colab, it told me none were available so I thought one has to pay for it...

Add support for VB-HMM clustering

cc @YannisTevissen ;-)

Empty ValueError raised during augmentation

Most likely an off-by-one rounding error due to https://github.com/asteroid-team/torch-audiomentations/blob/1590056ba3bbb06a75e24fb4bcec0fd10569a083/torch_audiomentations/utils/io.py#L232 and https://github.com/asteroid-team/torch-audiomentations/blob/1590056ba3bbb06a75e24fb4bcec0fd10569a083/torch_audiomentations/utils/io.py#L238 I believe using `math.floor` instead of `round` should fix the issue and we could safely remove the check that...