Closed yangye19960922 closed 4 years ago
This can be solved by running this file on cpu instead of GPU getting the gradient for RNNs without training is currently not supported by cuda
thanks for your reply,i add "model.train()"into demo, and it works
---Original--- From: "ayaabdelsalam91"<notifications@github.com> Date: Thu, Jun 11, 2020 16:49 PM To: "ayaabdelsalam91/Input-Cell-Attention"<Input-Cell-Attention@noreply.github.com>; Cc: "Author"<author@noreply.github.com>;"yangye19960922"<921629128@qq.com>; Subject: Re: [ayaabdelsalam91/Input-Cell-Attention] hello. something wrong when i run this project (#1)
This can be solved by running this file on cpu instead of GPU getting the gradient for RNNs without training is currently not supported by cuda
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
This will cause the weight of the model to update which is not something we want at this point
This will cause the weight of the model to update which is not something we want at this point
so you mean i need to run this file on cpu instead of adding 'model.train()' into file, becauce this operation will lead to wrong results?
yes
Hi, when i finish running trainModels.py and try to run saliency.py, an error occurs, it shows that RuntimeError: cudnn RNN backward can only be called in training mode could you please teach me how to solve this problem ?