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: Allow autodeps to auto add `fbcode//torchrec/inference:batching` to the buck target when we imported `torchrec/inference/Batching.h` Reviewed By: houseroad Differential Revision: D54284062

CLA Signed
fb-exported

Summary: Implement regression AUC metrics. Regression AUC is an extension of classification AUC. See Section 4.1.1 in https://arxiv.org/ftp/arxiv/papers/1205/1205.2618.pdf for related discussions. On a high level, regression AUC is an extension...

CLA Signed
fb-exported

Summary: Address OSS error by explicitly convert to numpy array and also using p50 for the median {F1461949323} Differential Revision: D54268961

CLA Signed
fb-exported

Summary: Dynamo has some gaps in support of generators, list comprehension etc. Avoiding them for now with syntactic changes Previous diff was reverted because recat was created on the target...

CLA Signed
fb-exported

Summary: Introduce BoundsCheckMode fused_param for TBE BoundsCheckMode. There is no reason really to run bounds_check_indices during inference use case (AIMP has it off by default: https://fburl.com/code/q8zhundg), and it causes issues...

CLA Signed

Summary: With _unwrap_kjt_for_cpu, and recording the runtime_device in QuantEmbeddingBag, the conditional for the input device in _unwrap_kjt is no longer needed, as this path should always record the cuda device...

CLA Signed

Summary: Title Differential Revision: D56688411

CLA Signed
fb-exported

Summary: Benchmark existing training benchmarks, training performance and memory on multi-gpu setups TrainPipelineBase | Runtime (P90): 13.1 s | Memory (P90): 8.4 GB TrainPipelineSparseDist | Runtime (P90): 12.7 s |...

CLA Signed

Summary: Provide optional support for non-determinstic behavor for more representative benchmarking Differential Revision: D56690894

CLA Signed