torchrec
torchrec copied to clipboard
pytorch
Pytorch domain library for recommendation systems
Differential Revision: D60993179
We aim to share embeddings between sparse features and sequence sparse features. For sparse features, we use `EmbeddingBagCollection`, and for sequence sparse features, we use `EmbeddingCollection`. Could you advise on...
Hello, it looks like EmbeddingBagCollection forces data type to be float32 or float16 during initialization. https://github.com/pytorch/torchrec/blob/main/torchrec/modules/embedding_modules.py#L179 Is there any support to make embedding be float8? Note, this is for training....
Summary: Fixes T192448049. The module call form an unusal call stack for the nodes: https://www.internalfb.com/phabricator/paste/view/P1507230978. This is currently not supported by unflattener and need some extra design to make it...
in the forward pass, in the table wise sharding, when pooling is executed? is it after alltoall communication? and executed on trainer local? where can I see the exact code...
Summary: When torch.exporting, device specializations can occur: https://fb.workplace.com/groups/1075192433118967/posts/1474166129888260/?comment_id=1474191496552390&reply_comment_id=1474683806503159 such as here: https://fburl.com/code/94ta7omp. Currently, the best solution is to do another pass of the graph and modify the device accordingly when...
I am trying to apply DATA_PARALLEL on the small embedding tables and it can work in EmbeddingBagCollection. However, when it comes to FusedEmbeddingBagCollection, it doesn't work and gets an error...
Summary: To improve inference, we want to make creating a KJT as cheap as possible, which means the init method is nothing more than a attribute setter. All other fields...