search

karpathy / pytorch-normalizing-flows

Normalizing flows in PyTorch. Current intended use is education not production.
845 stars 98 forks source link

Loss becomes NAN with the Glow model #9

Open ashispati opened 4 years ago

ashispati commented 4 years ago

The loss for the Glow model becomes NAN (sometimes). I have only tested it with the half-moons dataset with the default model in the Jupyter notebook. Did not face the same issue with the other models.

Screen Shot 2020-07-09 at 10 51 48 AM Screen Shot 2020-07-09 at 10 51 19 AM
YeonwooSung commented 4 years ago

For NaN values, you may try decreasing batch size.

richardrl commented 3 years ago

@YeonwooSung Why? Also, this issue seems to happen often