3 ms·
> I am comfortable with autograd/computation graphs, PyTorch, "classic" neural nets and ones used for vision-type applications, as well as the basics of Transfo
by optbuild 4y ago
> I am comfortable with autograd/computation graphs, PyTorch, "classic" neural nets and ones used for vision-type applications, as well as the basics of Transformer networks (I've trained a few smaller ones myself) and RNNs.
Maybe I am a bit off the track. But how do someone reach this state?
- newsoul 4y agoYou can try out fast.ai's courses. Also there is a youtube series by Andrej Karpathy that goes into much more detail and builds a library from scratch.