search

ParikhKadam / bidaf-keras

Bidirectional Attention Flow for Machine Comprehension implemented in Keras 2
GNU General Public License v3.0
64 stars 21 forks source link

Run only in CPU version instead of GPU version. #5

Closed hariteja97 closed 5 years ago

hariteja97 commented 5 years ago

The code that was written has GPU support but I want to run the code without GPU support i.e i need the code in CPU only. I tried to remove the multi-model-GPU package and its connections in the code but still I'M facing the issue

Kindly help me to convert the code to CPU support only.

ParikhKadam commented 5 years ago

Install tensorflow==1.12.0 instead of tensorflow-gpu==1.12.0. That will solve the problem I think. Or you can just reinstall the package by using pip install -U bidaf-keras[cpu].

Try any of these methods. If they don't work, please report here with the errors.. Thank you..

hariteja97 commented 5 years ago

@ParikhKadam Thanks a lot for the help and I have got a new doubt . I have tried this command

python3 -m bidaf --model_name=bidaf_50.h5 --do_lowercase train --epochs=1 --steps_per_epoch=1 --validation_steps=1

when I ran this command I got the error Capture

I dont understand what is this error Please help me it would be of great help

ParikhKadam commented 5 years ago

Have you downloaded our pretrained model "bidaf_50.h5" file? If yes, place the file in the path showed in this error which is "C:\Python36..." and the error will be solved..

hariteja97 commented 5 years ago

Yes parikh I have tried once check the error I got when I did that .. once check it sir .

Thanks a lot

On Fri, 17 May 2019, 16:08 Kadam Parikh, notifications@github.com wrote:

Have you downloaded our pretrained model "bidaf_50.h5" file? If yes, place the file in the path showed in this error which is "C:\Python36..." and the error will be solved..

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ParikhKadam/bidaf-keras/issues/5?email_source=notifications&email_token=AG7EBMELYR5WIPLBVT3YKETPV2DL3A5CNFSM4HMXQP5KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODVUNDUY#issuecomment-493408723, or mute the thread https://github.com/notifications/unsubscribe-auth/AG7EBMGSUDZZH52ERLRGLWLPV2DL3ANCNFSM4HMXQP5A .

ParikhKadam commented 5 years ago

The error clearly says that it cannot find the file "bidaf_50.h5" in that given path. Can you please show your files in that folder called saved_items?

hariteja97 commented 5 years ago

@ParikhKadam/bidaf-keras reply@reply.github.com Thanks a lot for ur help. I got the solution. I have a small issue here in my ubuntu system. Its like , in my ubuntu all packages are not install and I am not having any root access to connect to the ubuntu servers to download them as they are blocked and cannot be accessed.

So when I run the same code it went in theano backened instead of tensorflow. nd I was getting the error as in tensor( it is considering it as a list) has not attribute type and it stops

I got the dev1.1json and train1.1 json file and model everything till preprocessing is done . but training is not done.

Kindly help me in this issue

Thanks & Regards Hari Teja.N

On Fri, 17 May 2019, 23:09 Kadam Parikh, notifications@github.com wrote:

The error clearly says that it cannot find the file "bidaf_50.h5" in that given path. Can you please show your files in that folder called saved_items?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ParikhKadam/bidaf-keras/issues/5?email_source=notifications&email_token=AG7EBMAADENMQMZT4JRE3YLPV3UURA5CNFSM4HMXQP5KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODVVMVRI#issuecomment-493537989, or mute the thread https://github.com/notifications/unsubscribe-auth/AG7EBMBOQQG6LGXAIW4IK7TPV3UURANCNFSM4HMXQP5A .

ParikhKadam commented 5 years ago

Great to hear that it solved.. BTW, the new issue you mentioned isn't clear to me. I understood that you cannot install packages on your linux machine and hence, Keras uses theano as backend which throws such error.

It should not happen.. Changing the backend in Keras shouldn't throw errors. That's what Keras is meant to do, right?

What isn't clear to me is, is this an error in our project or Keras? Can you please check it and tell?

I am closing this thread now. If you think that the error is in this project, kindly open a new issue specific to that error and I will surely help..

Thank you..