torchrec
torchrec copied to clipboard
Benchmarking
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 | Memory (P90): 8.8 GB
Differential Revision: D56690925