Change loss function code

para_loss = F.binary_cross_entropy_with_logits(para_outputs.view(-1), para_labels, reduction='mean') / args.batch_size and sts_loss = F.mse_loss(sts_outputs.squeeze(), sts_labels.view(-1).float(), reduction='mean') / args.batch_size delete args.batch_size from each