junwenbai / mpvae Goto Github PK

Hi,

I notice that in training and testing the sampling times are very different. May I ask why? Can I use 10000 for training?

Cheers

Thanks for your nice work.I have some questions about Which method do you use to turn text into features one hot?If it is convenient for you, could you please provide your processing code?

Thanks for your nice work.I have some questions about data constructed by myself.I construct the reuters dataset as the way as the LaMP. when I run the model ,some metric and loss is Nan.But when i used the reuters dataset offered by[wenting-zhao/mavae](wget https://www.cs.cornell.edu/~junwen/data/MPVAE_data.tar.gz) ,the run result is normal.Could you please provide the code you used to construct the dataset？

（Pytorch Version）Thanks for your nice work.I have some questions about the specific config about different datasets.For example,when I run the model on the delicious data,some metric is Nan.I consider it maybe the inappropriate parameter I used. So，could you provide all configs for all datasets you used in the paper.Thanks a lot.