run `htop` and monitor the RAM usage to check
# neural-forecast
d
run
htop
and monitor the RAM usage to check
i
I've tried that, but I'm running on AWS so I'm seeing some funkiness.
d
Maybe reduce the batch size to 1 to see if that fixes it?
i
Do you know if the predict step is using the training or validation batch size?
looks like valid batch size, but I just started looking at the code yesterday so idk if that's correct
I also don't know how that effects the validation step