search

toshas / torch-fidelity

High-fidelity performance metrics for generative models in PyTorch
Other
1.01k stars 66 forks source link

FID between train_set and val_set of cifar10 is around 40, is this normal? #32

Closed dongzhuoyao closed 2 years ago

dongzhuoyao commented 2 years ago

Just want to use it to test the evaluation code.

Afaik, the current SOTA of cifar10 is around 3.