songwaimai

> The loss of mine in the training set is fine, while in cv set, some batches are nan. > It's definitely not gradient explosion. I don't know how to...

I also met this problem in generation task. I replaced the con 3x3 by involution, the loss in nan or inf.

> > 我在代任务中也遇到了这个问题。我将con3x3替换为对合，在NaN或INF中的损失。 > > 我也没解决，所以我已经快要放弃使用involution了 I also tried the gradient clipping method, but the NAN problem is not be solved, i will try to find some else methods which...

> > > > 我在代任务中也遇到了这个问题。我将con3x3替换为对合，在NaN或INF中的损失。 > > > > > > > > > 我也没解决，所以我已经快要放弃使用involution了 > > > I also tried the gradient clipping method, but the NAN problem is...

> > I replace the standard conv with involution and added bn, then the loss seems normal.But the final result is worse than edsr baseline with bn layer,even though i...