Training loss is decreasing while validation loss is NaN

abhijitjadhav1998 / Deepfake_detection_using_deep_learning

This projects aims in detection of video deepfakes using deep learning techniques like RestNext and LSTM. We have achived deepfake detection by using transfer learning where the pretrained RestNext CNN is used to obtain a feature vector, further the LSTM layer is trained using the features. For more details follow the documentaion.

GNU General Public License v3.0

397 stars 140 forks source link

Training loss is decreasing while validation loss is NaN #11

Closed Priyanshu999 closed 1 year ago

Priyanshu999 commented 1 year ago

I found that although the training loss is consistently decreasing, the validation loss remains as NaN (validation accuracy is constant after every epoch). Can you fix this?

abhijitjadhav1998 commented 1 year ago

Please debug the code. there seem to issue with gradient decent

Mithoon278 commented 5 months ago

did the issue solved?