Piotr Żelasko
Piotr Żelasko
Hmm, now I remember that’s why I hardcoded -1000 in the feature mixer in the first place. I am a bit afraid that changing -23 (which is log(energy_floor) for a...
Please update your Lhotse version: ```bash $ lhotse download libritts --dataset-parts dev-clean test-libritts Downloading LibriTTS parts: 0%| | 0/1 [00:00
How about using the custom field to hold all of the style variants? You can see how it could work with the following snippet: ```python import random import torch.utils.data from...
BTW torch.testing.assert_close prints out better diagnostic info (how many elements mismatch, by how much, etc)
Please change `assert torch.testing.assert_close(` to `torch.testing.assert_close(`, the actual assertion is inside, then you'll see more info
Ooops, not helpful with a single element tensor 🙈
Thanks for debugging this. I committed the relevant fixes in https://github.com/lhotse-speech/lhotse/pull/1355
Yea Lhotse should support these cases out of the box. Could you modify it?
Thanks, I’d love to help here. How do you envision stereo support? Do you think that the limiter needs to process 2-channel information together to estimate the gain, or is...
OK, I'll try to look into it in the following days.