Skip to content

About validation datasets #36

Description

@YYX666660

Hi~
The repository doesn't seem to have posted instructions on how to build the validation set. In the default apollo.yaml, the training num_samples is set to 40000.
So what is a reasonable size for the validation set?
When preparing the validation set offline, is it also necessary to use MusdbMoisesdbDataset to generate ori_wav/codec_wav pairs? (keep the parameter settings consistent with the training)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions