Automated Overtraining Detection (AOD)

working on stable realease of aod

currently latest version is rod 1.0.1 beta

Simple program that will automate searching the most optimal epoch for your RVC model

Installing AOD

1. Ensure you have Python installed.
2. Go to the Releases tab.
3. Download the latest version of AOD (aod.zip).

Unpack the Files

Unzip the downloaded aod.zip file to a location of your choice.

Local Training:

• If you plan to train locally, create a directory named AOD in your "Mangio-RVC".
• Drop the unpacked files inside the created AOD directory.

Colab Training:

• If you're using Google Colab, simply unzip the files and keep them in the same directory.

Install Requirements

Run the install-requirements.bat script to install the necessary dependencies.

Creating Configuration File

Use the config-editor.bat script to create and edit a configuration file according to your needs.

Personalizing Your Config

Model Name serves as the identifier for monitoring on the specified model.

Iterations Iterations determine the number of epochs that should be after reaching lowest point on the graph. For instance, if you select 100 iterations and there are only 20 epochs after lowest point, it will refresh after specified number of seconds.

IF YOU ALREADY HAVE TRAINED MODEL SET ITERATIONS TO 1

Refresh Rate The refresh rate indicates how frequently the program will update itself in seconds, if there is no value provided.

Environment Specifies your current workspace, offering two options: 'local' or 'colab.' When utilizing 'colab,' you will need to input specific links to the 'events.out.tfevents.0' files stored in your Google Drive. Which can be found by going to these directories:

• RVC_Backup
• YOUR MODEL NAME

Additionally, ensure that permissions are configured to allow anyone with the link to download these files.

Launch Tensorboard

Start Tensorboard by running the launch-tensorboard.bat script.

Run aod.bat

execute the aod.bat script, and get the outcome of sufficient epochs number.