Predict with new paraphrase examples about pqg-pytorch

[dev-chauhan]

You just have to make a custom dataset in PyTorch just like in dataloader.py if you are trying with a different paraphrase dataset. You can see in train_*.py files that show how to evaluate the model or how to get predicted paraphrase.

[srinivasraman18]

Hi @dev-chauhan ,
Can you pls tell me if it is possible to use a different test dataset on your pretrained model or do i need to retrain the model?
Thank you.

[ak-kumar-v]

You have to train the model first, the model will be saved in a directory "save" with .pt extension(have to create this, as mentioned in the README), then use checkpoint to load model and evaluate your dataset. I tried it, but output doesn't work well. If you get any solution kindly post here too