search

flavioschneider / master-thesis

26 stars 1 forks source link

question: Upsampler (238M) #1

Open playdasegunda opened 1 year ago

playdasegunda commented 1 year ago

Could you please share the code for the "Upsampler (238M)" model, which you found at this link, so that we can "upsample" a low-quality audio file using a repository, such as huggingface or colab?

Thank you, Lucas Rodrigues.

flavioschneider commented 1 year ago

The code for the upsampler can be found here github.com/archinet/audio-diffusion-pytorch, the config is the following:

    in_channels: 2
    upsample_factor: 16
    channels: [8, 32, 64, 128, 256, 512, 512, 1024, 1024]
    factors: [1, 4, 4, 4, 2, 2, 2, 2, 2]
    items: [1, 2, 2, 2, 2, 2, 2, 4, 4]
    attentions: [0, 0, 0, 0, 0, 1, 1, 1, 1]
    attention_heads: 12
    attention_features: 64

You have to train your own for now, I might release a pretrained version in the future.