pytorch

8c2e5699 - [PT2D] Make the speedup benchmark works with DDP + CompiledAutograd (#120454)

[PT2D] Make the speedup benchmark works with DDP + CompiledAutograd (#120454) With DDP + CompiledAutograd, we could not use the same parallelized model to do the test. This PR copies the model. Differential Revision: [D54094257](https://our.internmc.facebook.com/intern/diff/D54094257/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/120454 Approved by: https://github.com/yf225, https://github.com/xmfan