zxymark221
Results
1
comments of
zxymark221
🤝 I also noticed this issue today. Without dividing $\sigma$, the gradient estimation is around 100x smaller in magnitude (since sigma is usually at 0.01 magnitude). That explains why in...