torchaudio.models.squim_subjective_model

torchaudio.models.squim_subjective_model(ssl_type: str, feat_dim: int, proj_dim: int, att_dim: int) → SquimSubjective

Build a custome torchaudio.prototype.models.SquimSubjective model.

Parameters:

• ssl_type (str) – Type of self-supervised learning (SSL) models. Must be one of [“wav2vec2_base”, “wav2vec2_large”].

• feat_dim (int) – Feature dimension of the SSL feature representation.

• proj_dim (int) – Output dimension of projection layer.

• att_dim (int) – Dimension of attention scores.