search

nttcslab / byol-a

BYOL for Audio: Self-Supervised Learning for General-Purpose Audio Representation
https://arxiv.org/abs/2103.06695
Other
205 stars 35 forks source link

Model parameters cannot be trainable once they become requires_grad=False #7

Closed daisukelab closed 3 years ago

daisukelab commented 3 years ago

https://github.com/nttcslab/byol-a/blob/master/byol_a/models.py#L42

if p.requires_grad has to be removed.