pytorch

d776dd04 - perf(optim/dynamo): shortcut `is_sparse` iteration in SGD multi_tensor (#110648)

perf(optim/dynamo): shortcut `is_sparse` iteration in SGD multi_tensor (#110648) Originated: https://github.com/pytorch/pytorch/pull/110353#discussion_r1347806922 Speeds up significantly in non-sparse path (majority use-case). Benchmarks: https://github.com/pytorch/pytorch/issues/110506#issuecomment-1747732478 CC: @janeyx99 Pull Request resolved: https://github.com/pytorch/pytorch/pull/110648 Approved by: https://github.com/janeyx99