You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi @Orca-bit thanks for your interest for HugeCTR!
HugeCTR/src/embeddings and HugeCTR/embedding are two seperate embedding implementations. Compared with HugeCTR/src/embeddings, HugeCTR/embedding provides more flexible API for user to configure sharding.
To be more specific, HugeCTR/src/embeddings only support localized or distributed embedding but in HugeCTR/embedding user can specify the table placement for each GPU. If you are interested in the detail, you can refer to our paper published in RecSys 2024 entitled Embedding Optimization for Training Large-scale Deep Learning Recommendation Systems with EMBark.
Our goal is to improve HugeCTR/embedding and will deprecate HugeCTR/src/embeddings in the future.
This template is for generic questions that a user may have in using HugeCTR.
Note: Before filing an issue, you may want to check out our compiled Q&A list first.
The text was updated successfully, but these errors were encountered: