We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Going to look into this in detail now, but got the following error while running the imagenet ipynb
RuntimeErrorTraceback (most recent call last) <ipython-input-17-c4f537643c1f> in <module>() 8 prob_outputs_dog = Variable(torch.zeros(1,1000)) ; prob_outputs_dog.data[:,dog_id] += 1 9 ---> 10 prob_inputs_cat = eb.excitation_backprop(model, inputs, prob_outputs_cat, contrastive=False) 11 prob_inputs_dog = eb.excitation_backprop(model, inputs, prob_outputs_dog, contrastive=False) /home/rrawat/coderepo/excitationbp/excitationbp/utils.py in excitation_backprop(model, inputs, prob_outputs, contrastive, target_layer) 45 if not contrastive: 46 outputs = model(inputs) ---> 47 return torch.autograd.grad(top_h_, target_h_, grad_outputs=prob_outputs)[0] 48 49 pos_evidence = torch.autograd.grad(top_h_, contr_h_, grad_outputs=prob_outputs.clone())[0] /home/rrawat/anaconda2/lib/python2.7/site-packages/torch/autograd/__init__.pyc in grad(outputs, inputs, grad_outputs, retain_graph, create_graph, only_inputs) 151 return Variable._execution_engine.run_backward( 152 outputs, grad_outputs, retain_graph, --> 153 inputs, only_inputs) 154 155 if not torch._C._autograd_init(): /home/rrawat/anaconda2/lib/python2.7/site-packages/torch/autograd/function.pyc in apply(self, *args) 89 90 def apply(self, *args): ---> 91 return self._forward_cls.backward(self, *args) 92 93 /home/rrawat/coderepo/excitationbp/excitationbp/functions/eb_linear.pyc in backward(ctx, grad_output) 28 29 input.data = input.data - input.data.min() if input.data.min() < 0 else input.data ---> 30 grad_output /= input.mm(weight.t()).abs() + 1e-10 # normalize 31 ### stop EB-SPECIFIC CODE ### 32 /home/rrawat/anaconda2/lib/python2.7/site-packages/torch/autograd/variable.pyc in __idiv__(self, other) 847 848 def __idiv__(self, other): --> 849 return self.div_(other) 850 851 def __pow__(self, other): /home/rrawat/anaconda2/lib/python2.7/site-packages/torch/autograd/variable.pyc in div_(self, other) 357 if not isinstance(other, Variable) and not torch.is_tensor(other): 358 return DivConstant.apply(self, other, True) --> 359 raise RuntimeError("div_ only supports scalar multiplication") 360 361 def pow(self, other): RuntimeError: div_ only supports scalar multiplication
The text was updated successfully, but these errors were encountered:
Interesting. I just tried to reproduce your error with PyTorch 0.3 and the latest commit of excitationbp. I wasn't able to reproduce your error.
I did push a small two-line update to this repo about an hour ago, but I'm not sure how that could have caused the error you are getting.
Sorry, something went wrong.
No branches or pull requests
Going to look into this in detail now, but got the following error while running the imagenet ipynb
The text was updated successfully, but these errors were encountered: