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

Summary: We are enabling custom perf models while evaluating candidate plans in `MemoryBalancedPartitioner`. This will enable us to leverage the new critical path model defined in `NoopCriticalPathPerfModel` while trying to...

CLA Signed
fb-exported

Summary: Update the critical path definition in the planner logs to match what we think is the most realistic option. See the docstring for the detailed modeling. Differential Revision: D73207877

CLA Signed
fb-exported

Summary: Original commit changeset: 5ac77088cf97 Original Phabricator Diff: D72010614 Issues in `ValueError: Unexpected keyword arguments: `fuse_state_tensors` Differential Revision: D73203584

CLA Signed
fb-exported

Summary: Bucket offsets and sizes in torchrec shard metadata for bucket wise sharding for ZCH v.Next Differential Revision: D72921209

CLA Signed
fb-exported

Differential Revision: D72702173

CLA Signed
fb-exported

Summary: This is to upload the torchrec RFC to open source repository. Differential Revision: D72895838

CLA Signed
fb-exported

Differential Revision: D72871677

CLA Signed
fb-exported

Summary: `stride_per_key_per_rank` should be a variable whose value is dynamically after input_dist. Updating its type to `Union[Optional[torch.Tensor], Optional[List[List[int]]]]` to be backward compatible. Differential Revision: D72658640

CLA Signed
fb-exported

…the code and spec in DCN V1 paper The original comment mentioned the bias was scaled by x_0. The paper mentions the bias is a separate term in the update...