tbh-98/reproducing-of-cpm-nets-cross-partial-multi-view-networks

Code in pytorch for reproducing experiments in "CPM-Nets: Cross Partial Multi-View Networks"

reproducing-of-cpm-nets-cross-partial-multi-view-networks's Issues

Question

Do you have time？

CPM_Nets.py报错

你好~
CPM_Nets.py文件中97行，出现如下错误，请问要怎么处理呢？
line 97, in train
Reconstruction_LOSS.backward(retain_graph=True)
File "/home/ply/anaconda3/lib/python3.7/site-packages/torch/tensor.py", line 245, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "/home/ply/anaconda3/lib/python3.7/site-packages/torch/autograd/init.py", line 147, in backward
allow_unreachable=True, accumulate_grad=True) # allow_unreachable flag
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [150, 216]], which is output 0 of TBackward, is at version 3; expected version 2 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
网上查了一圈说是计算梯度的时候+=替换之类的问题，但我看代码中，反向传播时就只有retrain_graph=True，重构损失函数中loss也是loss=loss+....的表示，不知道如何解决。

关于编码网络

你好，编码网络部分代码中是直接潜在表示和权重进行了相乘，并没有看到需要经过激活函数，是这样吗？

I want to contact to you

Hello,I want to communicate with you ,can I get your email