Retrieval-based-Voice-Conversion-WebUI icon indicating copy to clipboard operation
Retrieval-based-Voice-Conversion-WebUI copied to clipboard

Published 20 hours ago • verified publisher icon fumiama

Training restarts at checkpoint instead of continuing

Open TheTrustedComputer opened this issue 5 months ago • 1 comments

Describe the bug This seems to be an unintentional regression. If one stops training mid-session and resumes, it starts from the beginning instead of continuing with the saved D and G checkpoints. The initial output shows these files aren't even loaded when they exist in the file system.

To Reproduce Create any RVC model in the training tab, stop after the first checkpoint, and repeat, using the same parameters.

Expected behavior Training should continue from the previously saved checkpoint, as RVC always has.

Screenshots N/A

Desktop (please complete the following information):

  • OS and version: Arch Linux
  • Python version: 3.10.13
  • Commit/Tag with the issue: main

Additional context This is likely due to the commit 277eadca9ad25b9092aaccb2144b05fae0e190af, which renamed checkpoint files from "2333333" to "latest". Resumption from the latest checkpoint worked in an older commit, but it's broken now.

TheTrustedComputer avatar Sep 20 '24 06:09 TheTrustedComputer