search

Justin1904 / TensorFusionNetworks

Pytorch Implementation of Tensor Fusion Networks for multimodal sentiment analysis.
172 stars 44 forks source link

Weights of audio sub network returning NAN #6

Closed purvanshi closed 6 years ago

purvanshi commented 6 years ago

The layer1 weights of the audio subnetwork start giving NAN values. Particularly talking about the output of self.linear_1(dropped) I tried adjusting the learning rate but it doesn't help.

Justin1904 commented 6 years ago

Are you using your own data or the datasets from CMU Multimodal SDK? Have you examined if there're NaN values in the input? For the past version of SDK I wrote code for checking and removing inf and nan values manually, it might be the case that in your data or the new version of SDK's data there're also inf or nan values in the input. In my previous experience, this happens in particular with COVAREP acoustic features.

purvanshi commented 6 years ago

yes the problem is with acoustic features. The input is nan at several points. Thanks for the help.