Multinomial VAE - performance - Githubissues

recommenders-team / recommenders

Best Practices on Recommendation Systems

https://recommenders-team.github.io/recommenders/intro.html

MIT License

19.28k stars 3.11k forks source link

Multinomial VAE - performance #1249

Open PaulCristina opened 3 years ago

PaulCristina commented 3 years ago

I've noticed some differences in training time and performance between this tensorflow 2 implementation and the original version:

lower ratings should be removed

user-to-movie interactions with rating <=3.5 are filtered out but they are used to generate test_data_te_ratings, val_data_te_ratings that are used to compute the metrics for the model. Could I ask why?
huge spike in memory consumption, using dataset 20m when the training is started will gobble up ~25gb RAM while the initial version will only use ~4gb RAM. Is this is a bug?

Thank you

EvgeniaChroni commented 3 years ago

Regarding the lower ratings, it is proposed by the examined paper Section 4.1