pytorch

b946e7a7 - [TorchArrow][efficiency][2/n] added inference_wrapper_run_flat_out fused version registration call (#81131)

[TorchArrow][efficiency][2/n] added inference_wrapper_run_flat_out fused version registration call (#81131) Summary: Added registration call to make pytorch runtime aware of `inference_wrapper_run_flat_out` functionality. Added fused version of op which will enable out variants when [OptimizeGraph pass](https://fburl.com/code/lsagnmge) is made. Next: need to add variadic version of fused and unfused op. Reviewed By: tgalkovskyi Differential Revision: D37139204 Pull Request resolved: https://github.com/pytorch/pytorch/pull/81131 Approved by: https://github.com/tenpercent, https://github.com/qxy11