Piotr Żelasko
Piotr Żelasko
Does pdb work with seg faults to display python stack traces? Also I recall gdb can be made to display python stack traces but I don’t remember how.
Lazy loading uses orjson if available for faster json parsing. Can you try to install/uninstall it and see if it changes anything?
Yes, please check out the tips here https://lhotse.readthedocs.io/en/latest/corpus.html#adding-new-corpora and feel free to use any of the existing recipes as a starting point https://github.com/lhotse-speech/lhotse/tree/master/lhotse/recipes
Hmm I'm not sure what went wrong, but you may be able to work around the issue by using the resampling option in first ffmpeg call (`-ar 8000` instead of...
Otherwise you are encouraged to use resampling via Lhotse API (e.g. `cuts = cuts.resample(8000).resample(16000)`, also works with `recordings`)
I think it should be straightforward to modify the code to iterate over every channel in a recording, create supervisions specifically for that channel, and bind it all together into...
CC maybe @desh2608 would also be interested
I never optimized it for speed. There is a lot of discussion about it in the Whisper repo, maybe you can find something from there useful. I'd be happy to...
Good idea, do you mind submitting a PR?
That's smart! As a side note I was also considering if we should support more nuanced padding strategies, e.g. with mean example/batch value (so that it doesn't influence things such...