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

Errors in tensorflow while training the model #9

Closed hariteja97 closed 5 years ago

hariteja97 commented 5 years ago

Whenever I try to train the model.. I get this error and I don't know how to rectify this error. IMG20190522140118

This is the error I get..

Thanks in advance.

ParikhKadam commented 5 years ago

Which version of tensorflow are you using?

hariteja97 commented 5 years ago

I am using tensorflow 1.13.1 and Keras as 2.2.4.. and it's python 3.6.8 version..

On Wed, 22 May 2019, 19:42 Kadam Parikh, notifications@github.com wrote:

Which version of tensorflow are you using?

— 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/9?email_source=notifications&email_token=AG7EBMARCOS6AAALJ574NBTPWVIGVA5CNFSM4HOR5MYKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODV7F3DA#issuecomment-494820748, or mute the thread https://github.com/notifications/unsubscribe-auth/AG7EBMBCOZSKKTLE63LM4NTPWVIGVANCNFSM4HOR5MYA .

ParikhKadam commented 5 years ago

Try using tensorflow 1.12.0 and report is the error persists.

I am going to update this project to support tensorflow 2.0 and use tf.keras instead of keras as soon as a stable version to tensorflow 2.0 releases..

ParikhKadam commented 5 years ago

@hariteja97 Did this solve the issue?

hariteja97 commented 5 years ago

@kadamparikh. Thanks a lot it helped me a lot ..I have just re installed 1.12.0 version tensorflow which solved my issue .but I was trying the same code with theano backend but it's not working .

It would be great if we can implement this in theano also

One more thing.

  1. How to make the model much more accurate like there are some exceptions with this code like few answers are not developed properly.. to make it more precise

    Can u explain how to use BERT model and or any other model for making it more accurate

On Tue, 28 May 2019, 11:33 Kadam Parikh, notifications@github.com wrote:

@hariteja97 https://github.com/hariteja97 Did this solve the issue?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ParikhKadam/bidaf-keras/issues/9?email_source=notifications&email_token=AG7EBMEANCHXVTIKH52GH7LPXTDLFA5CNFSM4HOR5MYKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWLBYVY#issuecomment-496376919, or mute the thread https://github.com/notifications/unsubscribe-auth/AG7EBMA6QCVRVUCUJS23MPDPXTDLFANCNFSM4HOR5MYA .

ParikhKadam commented 5 years ago

@hariteja97 The model doesn't work well with Theano is a problem with Keras. You should try opening an issue at Keras Repo with the errors you get.

Improving model's accuracy needs training. The models I provided aren't actually completely trained. And I don't have resources for further training. Try that yourself..

Thank you..