Where to obtain XTTS-v2 original_model_files (tokenizer, checkpoints, mel_stats) for local fine-tuning? #4357
Unanswered
adamcomiskey
asked this question in
General Q&A
Replies: 1 comment
-
|
Not sure where you saw that link, the correct Huggingface link is https://huggingface.co/coqui/XTTS-v2/tree/main Otherwise, if you ever ran the model locally, Coqui will have downloaded them as well into the default location: https://coqui-tts.readthedocs.io/en/latest/faq.html#where-does-coqui-store-downloaded-models |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Hi everyone,
Utter newbie here and a first time poster...... sorry in advance if this is a dumb question or I'm going about this incorrectly.
I’m setting up XTTS-v2 fine-tuning locally (Phase 2, decoder training) using the latest coqui-tts fork (v0.27.2 + coqui-tts-trainer 0.3.1).
The training script expects the folder
XTTS_v2.0_original_model_files
containing
config.json, encoder.pth, decoder.pth, dvae.pth, mel_stats.pth, and tokenizer.json.
However, all previous public download links such as
https://huggingface.co/coqui/XTTS-v2/tree/main/original_model_files
now return 404.
Could someone from the Coqui team (or anyone who has them cached) confirm whether these files are still available for open fine-tuning, and if so, where to obtain them safely?
Thanks in advance — I’d really like to continue from an existing Phase 1 encoder checkpoint rather than start from scratch.
Adam
Beta Was this translation helpful? Give feedback.
All reactions