Testimg把no-cuda修改为True之后仍然报cuda out of memory

[pengzha0]

理论上此时我用CPU训练了，但是仍然报Cuda 错误

[JiaRenChang]

Hi, @woshiniqiangdadebaba
您可以把nn.DataParallel也comment看看

[jangsooyoung]

change batch_size <===

ex)
TrainImgLoader = torch.utils.data.DataLoader(
DA.myImageFloder(all_left_img,all_right_img,all_left_disp, True),
batch_size= 2, shuffle= True, num_workers= 2, drop_last=False) #batch_size= 12, shuffle= True, num_workers= 8, drop_last=False)

TestImgLoader = torch.utils.data.DataLoader(
DA.myImageFloder(test_left_img,test_right_img,test_left_disp, False),
batch_size= 2, shuffle= False, num_workers= 2, drop_last=False) #batch_size= 8, shuffle= False, num_workers= 4, drop_last=False)