why do we need to define a l2_loss instead of directly using nn.MSELoss? - Githubissues

samet-akcay / ganomaly

GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training

MIT License

853 stars 213 forks source link

why do we need to define a l2_loss instead of directly using nn.MSELoss? #59

Open jianzhuwang opened 4 years ago

jianzhuwang commented 4 years ago

RT. Looking forward to your reply.