Repository navigation
|
Hi @pnnbao97, I've been following the project and I'm very impressed with the results from the VieNeu-TTS-1000h dataset. As I am interested in understanding how to properly structure and label data for future contributions (especially for the style/emotion tags we discussed), would it be possible for you to share a small sample or a detailed documentation of the dataset format used in pnnbao-ump/VieNeu-TTS-1000h? Specifically, I’m curious about:
Having a clear reference of your 1000h dataset's format would help the community prepare compatible data, ensuring that any future fine-tuning or data contribution aligns perfectly with your current pipeline. Thank you for your hard work and for supporting the Vietnamese TTS community! Best regards, |
Replies: 1 comment
|
Hi @chucly112, A reference dataset format is already publicly available here: https://huggingface.co/datasets/pnnbao-ump/VieNeu-TTS-140h The VieNeu-TTS-140h dataset follows the same structure used in training, including directory organization and metadata format. You can use it as a direct guideline for preparing compatible datasets or future community contributions. In short:
Hope this helps 👍 |
Hi @chucly112,
A reference dataset format is already publicly available here:
https://huggingface.co/datasets/pnnbao-ump/VieNeu-TTS-140h
The VieNeu-TTS-140h dataset follows the same structure used in training, including directory organization and metadata format. You can use it as a direct guideline for preparing compatible datasets or future community contributions.
In short:
Hope this helps 👍