Hervé BREDIN comments

Results 270 comments of


                                            Hervé BREDIN

Incompatibility between speaker diarization pipeline and Waveform preprocessor

Relevant: happens for AliMeeting (that contains 8-channel audio)

Training on a large set is much slower than on a smaller set - proportionally

Thanks Jan. Not quite sure. I was just (like 2 minutes ago) discussing with a colleague about the fact that pyannote is missing some kind of profiling to debug this...

Training on a large set is much slower than on a smaller set - proportionally

Let me know if/how I can help.

Training on a large set is much slower than on a smaller set - proportionally

It is a good idea to do your tests with the latest `develop` commit, indeed. However, know that it will not fix this issue. Abouth this line of code: ```python...

Training on a large set is much slower than on a smaller set - proportionally

Thanks Jan! Would be great indeed if you could run with SSD with the 3 versions: * `develop` * `develop` with `np.searchsorted` * `develop` with your `dict`-based version.

Training on a large set is much slower than on a smaller set - proportionally

Oh. I did not realize you were using such a small number of workers. I usually use 10 :)

Training on a large set is much slower than on a smaller set - proportionally

I'd be curious to have a look at your `dict`-based approach :) Regarding your last point (long epoch), you could actually use the `limit_train_batches` option of `pytorch-lightning`'s `Trainer` to reduce...

Training on a large set is much slower than on a smaller set - proportionally

I guess you need to fork the repo, push to your own fork, and open a PR from it?

Training on a large set is much slower than on a smaller set - proportionally

Re-opening it as I think it is worth looking into it (and also there's your related PR that I still need to have a look at).

Fix/update tutorials notebooks

Can you take care of the conflict in `add_your_own_task` notebook?