Closed shay86 closed 7 years ago
I think the truncation behavior of ClassNLLCriterion changed. What you might be able to do is to reshape the inputs to be the same size as the targets in https://github.com/DeepLearning4BioSeqText/Paper16-AAAI-MUST-CNN/blob/master/Trainer.lua#L70, and do the opposite at https://github.com/DeepLearning4BioSeqText/Paper16-AAAI-MUST-CNN/blob/master/Trainer.lua#L92.
Sorry we're not really keeping this up to date, I might do a pytorch version of this later. You can also try installing an older version of torch, from December 2015 or so.
thank you for answering me can you send me which version of Torch ,Lua and Cuda you installed and used .So I try to install it because i think the error not from the code it self but because it is not compatible with the new versions
I do not have a version installed anywhere anymore. I'm not completely sure if this version works (just look in torch7/distro and pick an old enough commit) but it should. I believe we were on CUDA 5, though CUDA 6 might also work.
git clone https://github.com/torch/distro.git ~/torch --recursive
cd ~/torch; git checkout b3c2671b0fb7b19a5e3a92e1f5e494a8c3f2fcc9
bash install-deps;
./install.sh
hi, I got this new issue and I am new to torch i don't know how to fix it. i Google it and it seem to be some common version issue. Do any on know how to solve this problem. Knowing I am using the project data without any change. thanks for help in advance.