search

kumar-shridhar / PyTorch-BayesianCNN

Bayesian Convolutional Neural Network with Variational Inference based on Bayes by Backprop in PyTorch.
MIT License
1.42k stars 323 forks source link

Error in loss.backward() #52

Closed ghost closed 4 years ago

ghost commented 4 years ago

Hi, I'm trying to adapt your code to a Regression problem. I'm getting the following error.

/pytorch/torch/csrc/autograd/python_anomaly_mode.cpp:57: UserWarning: Traceback of forward call that caused the error: File "bayesian_specnet.py", line 141, in run() File "bayesian_specnet.py", line 125, in run train_loss, train_kl = train_model(net, optimizer, criterion, train_loader, num_ens=train_ens, beta_type=beta_type, epoch=epoch, num_epochs=n_epochs) File "bayesian_specnet.py", line 41, in train_model net_out, _kl = net(inputs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in call result = self.forward(*input, *kwargs) File "/content/probabilistic-nn-cars/layers/misc.py", line 19, in forward x = module(x) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 541, in call result = self.forward(input, **kwargs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/activation.py", line 603, in forward return F.softplus(input, self.beta, self.threshold)

Traceback (most recent call last): File "bayesian_specnet.py", line 141, in run() File "bayesian_specnet.py", line 125, in run train_loss, train_kl = train_model(net, optimizer, criterion, train_loader, num_ens=train_ens, beta_type=beta_type, epoch=epoch, num_epochs=n_epochs) File "bayesian_specnet.py", line 53, in train_model loss.backward() File "/usr/local/lib/python3.6/dist-packages/torch/tensor.py", line 166, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph) File "/usr/local/lib/python3.6/dist-packages/torch/autograd/init.py", line 99, in backward allow_unreachable=True) # allow_unreachable flag RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [256, 16, 573]], which is output 0 of SoftplusBackward, is at version 1; expected version 0 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!

Is this a version-related error? Could anyone give a suggestion? Thanks