Crash during training: paramter logits has invalid values

[Sertingolix]

You could use something like this for your custom model:

RNN L2 weights regularization

You can get L2 loss with the custom loss API as follows. @override(ModelV2) def custom_loss(self, policy_loss, loss_inputs): l2_lambda = 0.01 l2_reg = torch.tensor(0.) for param in self.parameters(): l2_reg += torch.norm(param) self.l2_loss = l2_lambda * l2_reg assert self.l2_loss.requires_grad, "l2 loss no gradient" custom_loss = self.l2_loss # depending on input add los…

I hope that helps.