pytorch做目标检测，网络训练参数报错

Question

问题遇到的现象和发生背景

用的pytorch1.2，python3.7做一个目标检测的模型，上层函数里的参数和实际调用的参数一致，但是在训练时pytorch里面的moudle.py一直报错：

Traceback (most recent call last):
  File "train_TextGraph.py", line 237, in <module>
    main()
  File "train_TextGraph.py", line 218, in main
    train(model, train_loader, criterion, scheduler, optimizer, epoch, logger)
  File "train_TextGraph.py", line 75, in train
    output, gcn_data = model(img, gt_roi, to_device)
  File "/home/gpu/anaconda3/envs/DRRG/lib/python3.7/site-packages/torch/nn/modules/module.py", line 547, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/gpu/Desktop/DRRG-master/network/textnet.py", line 125, in forward
    gcn_pred = self.gcn_model(feat_batch, adj_batch, h1id_batch)
  File "/home/gpu/anaconda3/envs/DRRG/lib/python3.7/site-packages/torch/nn/modules/module.py", line 547, in __call__
    result = self.forward(*input, **kwargs)
TypeError: forward() takes 3 positional arguments but 4 were given

根据指示找到该函数，发现这个函数只有两句代码，请问这是什么原因，该咋解决？

 def forward(self, *input):
        raise NotImplementedError

Answer 1

你这个报错是参数设置问题，如果按照教程敲，大部分是版本更新问题，试试降低版本