Semi-supervised code pytorch

I have the data π‘‹βˆˆβ„π‘›Γ—π‘‘ with 𝑛=300,000 the number of data and 𝑑=2048 the number of variables. Of the 𝑛 given I only have about 10,000 labels π‘¦βˆˆ{0.1}} so for about 290,000 individuals, I don’t know the labels.

Is there a way to create a good classifier?

I think this problem is widely known as weakly or semi-supervised learning. I just can’t find adequate code on GitHub in pytorch , preferably quite simple and understandable.

Thank you in advance