wrk226 / pytorch-multimodal_sarcasm_detection Goto Github PK

Hello @wrk226 , I didn't find the raw image data in this repository. I wonder to know how to get it?

In LoadData.py,the 91st line of code, "image_feature_data" does not appear in the GitHub file

Hallo，thanks for your code, I find a parameter named fc_dropout_rate in train.ipynb, and it's used in 'FinalClassifier.ClassificationLayer(fc_dropout_rate)', and this parameter is definded in a list.
But it dont appeared in 'FuseAllFeature.FinalClassifier()', so, It's my personal negligence or some problem in code?

Hello，thanks for your code. In the code, a "text_embedding/vector.txt" file is used to represent a 175 sentence vector as a 75200 vector. How to use a vector to represent a word? Do I need to train a text_embedding file for word transformation？

It seems that little different from the original code by TF1, the origin code using the split train, valid and test set separately, but in this implement, the all set had been mixed and resplit. I try to use the three sets directly but it will make the model couldn't be trained. Sorry for the bother, but it really make me curious.

Hi, I am trying to run your code. But I have a problem in getting the attribute modality.
It seems that there is not the trained predictor in this repository.
Would you please explain how to train and which specific dataset and labels are used for training?
Thanks!