Still amazed that people are having a hard time with basic loss functions in PINN. How do you even implement supervised learning without this stuff? https://www.reddit.com/user/cae_shot