torchrec
torchrec copied to clipboard
pytorch
Pytorch domain library for recommendation systems
Summary: This diff reverts D76061842 This breaks the ifu_udd conveyor (models): https://www.internalfb.com/conveyor/mvai/video_ifu_udd_lsr/releases/951.1/nodes/cogwheel_blue_reels_vdd_trunk_metrics_test/runs/1450737805912090?pipeline_names[0]=main&node_details_tab=run_details_summary Depends on D76061842 Reviewed By: catalinii Differential Revision: D76268115
Summary: ### Major changes - Create a `mpzch` folder under the `torchrec/github/examples` folder - Implement a simple SparseArch module with a flag to switch between original and MPZCH managed collision...
Hi dear torchrec developers. I found a fatal bug when using EmbeddingCollection. The full stack is ``` [rank0]: File "/home/admin/hippo/worker/slave/aop_418921_aop_launcher_job_temp_m_20250528093245_6524584_job.worker_0_57_12/train/test_ebd.py", line 44, in [rank0]: main() [rank0]: File "/home/admin/hippo/worker/slave/aop_418921_aop_launcher_job_temp_m_20250528093245_6524584_job.worker_0_57_12/train/test_ebd.py", line 36,...
Does ManagedCollisionEmbeddingBagCollectionSharder only support cuda？ I want to use ManagedCollisionEmbeddingBagCollection in CPU only environment. But I cannot resolve the "RuntimeError: No available sharding type and compute kernel combination after applying...
Differential Revision: D76031572
I have no GPU
Hi there, the ManagedCollisionEmbeddingCollection with **multiple tables + shared features** returns all-zero embeddings after applying `apply_optimizer_in_backward` with **RowWiseAdagrad**. This result is unexpected. This bug likely relates to the initialization and...
I found an interesting phenomenon that could be enhanced when using KeyedJaggedTensor. ``` import torch from torchrec.sparse.jagged_tensor import JaggedTensor, KeyedJaggedTensor values = [ torch.Tensor([1.0]), torch.Tensor(), torch.Tensor([7.0, 8.0]), torch.Tensor([10.0, 11.0, 12.0]),...