3 ms·
> It 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
by snicker7 5y ago
> It 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/6a03e7b151c86b32d74b41af54d495c6y https://gist.github.com/ChrisRackauckas/6a03e7b151c86b32d74b...
- dklend122 5y agoYea, I meant parity for vanilla ML models. For anything off that beaten path it's much much faster