code/train.py
@@ -134,7 +134,7 @@ criterion = nn.CrossEntropyLoss().cuda() reg_criterion = nn.MSELoss().cuda() # Regression loss coefficient alpha = 0.00 alpha = args.alpha idx_tensor = [idx for idx in xrange(66)] idx_tensor = Variable(torch.FloatTensor(idx_tensor)).cuda(gpu)
