Comment on Growing open source from Torch to PyTorchparentComments−snicker75yIt had. It's now around parity with pytorch.In some cases, it is much faster.Consider Neural Stochastic Differential Equations, Flux is literally over 70,000x faster than Google's PyTorch-based implementation:https://gist.github.com/ChrisRackauckas/6a03e7b151c86b32d74b...−dklend1225yYea, I meant parity for vanilla ML models. For anything off that beaten path it's much much faster
Comments
In some cases, it is much faster.
Consider Neural Stochastic Differential Equations, Flux is literally over 70,000x faster than Google's PyTorch-based implementation:
https://gist.github.com/ChrisRackauckas/6a03e7b151c86b32d74b...
Yea, I meant parity for vanilla ML models. For anything off that beaten path it's much much faster