torchrec icon indicating copy to clipboard operation
torchrec copied to clipboard

Pytorch domain library for recommendation systems

Results 455 torchrec issues
Sort by recently updated
recently updated
newest added

Differential Revision: D53707387

CLA Signed
fb-exported

Summary: allow inverse indices to be included in new KJT after permute. For modifying vbe KJT outside of input dist. Differential Revision: D53732141

CLA Signed
fb-exported

Summary: Add new metric called ScalarMetric, which logs the tensor passed in as the label to tensorboard. This will help users of APS log custom scalers in production runs Differential...

CLA Signed
fb-exported

Summary: Performance optimization for KeyedJaggedTensor regrouping. Avoid unnessary temporary lists & dict by storing all keys in a pair of dicts Single dict allows us single look up instead on...

CLA Signed

Summary: X-link: https://github.com/pytorch/FBGEMM/pull/2310 ATT Differential Revision: D53364919

CLA Signed
fb-exported

Differential Revision: D53474222

CLA Signed
fb-exported

Summary: bdhirsh introduced a change in D51418076 where intermediate leafs with grad will cause a graph break. This leads to graph breaks in training our APS model. Example: P1156881935 The...

CLA Signed
fb-exported

Differential Revision: D53434630

CLA Signed
fb-exported

Summary: We currently do not support this functionality in embedding collection operations. However, there are emerging use cases where this will be more useful. Differential Revision: D52356440

CLA Signed
fb-exported

Summary: `EmbeddingCollection` and `EmbeddingBagCollection` contain duplicate code for methods `_pre_load_state_dict_hook()` and `_initialize_torch_state()`. Refactor this out to an abstract class `ShardedTensorEmbeddingModule` which both EC and EBC inherit Differential Revision: D53198210

CLA Signed
fb-exported