Create .keep in new logs/models folder to enhance sorting

[kubinka0505]

Description

Added .keep in new logs/models directory to enhance sorting for users that have many weights and not scroll in all logs directory

Existing glob pattern of .pth search already supports this

Motivation and Context

How has this been tested?

Screenshots (if appropriate):

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added tests to cover my changes.
• All new and existing tests passed.

[AznamirWoW]

How exactly does it enhance sorting?

[kubinka0505]

How exactly does it enhance sorting?

Currently models are loaded from logs folder, as stated in:

https://github.com/IAHispano/Applio/blob/main/tabs%2Finference%2Finference.py#L100

By adding a custom directory user can move existing models with their index files so structure may now look like:

logs/
    models/
        models_2/
            model.pth
            model.index
        model_2/
            model.pth
            model.index
        ...
    reference/
    mute/
    ...

[AznamirWoW]

How's having

logs\models\model_1
logs\models\model_2

provides better sorting than current

logs\model_1
logs\model_2

?

[kubinka0505]

How's having

logs\models\model_1 logs\models\model_2

provides better sorting than current

logs\model_1 logs\model_2

?

vs

Combine it with long model names and this is self-explanatory.

I haven't made this pull to make it move the trained model automatically to new directory, as it's agains the training continuation scheme - my change is purely cosmetic and could be classified as chore

Creating this directory will not break the training scheme as variables defined in this process are predefined, not "looked upon" (as per for x in os.listdir(): find x pseudologic)

[AznamirWoW]

okay, so it makes a neat view in Windows Explorer.
Except uses have no idea they need to place the models there, the download model function still saves models into logs\model_name, training uses logs\model_name, etc

[blaisewf]

Basically, we’re creating an empty folder that doesn’t save anything automatically (which will be more interesting)

we’re just letting know there’s an empty folder available that you can use to save models if needed

that said, it may not be strictly necessary, since you can create the folder yourself if you want.

[kubinka0505]

Imagine an edge case where I'd like to train or download a MODEL that is called mute which has mute.pth and mute.index inside archive.

Since that already exists in the logs directory, it wouldn't be downloaded or could, and most likely will, result in a conflict

If folder already existed, then it could be moved into that models directory... (not implemented)

[blaisewf]

yeah so this PR is missing all the real implementation to use this folder