725c327284
[nativert] add memory overlap debug assertion ( #157290 )
...
Summary: better safe than sorry. will throw if memory overlap detected when using planned tensors and debug mode is enabled -- this will make our planning unit tests more robust.
Test Plan:
ci
Rollback Plan:
Differential Revision: D77327841
Pull Request resolved: https://github.com/pytorch/pytorch/pull/157290
Approved by: https://github.com/SherlockNoMad , https://github.com/zhxchen17
2025-07-14 19:12:41 +00:00
018e9826a2
[nativert] hook up memory planning to execution frame ( #157053 )
...
Summary: pretty simple. if planner exists, which implies that planning is enabled, create a manager for each frame. the associated serial executor will use the withMemoryPlannner fn to ensure the deallocation is done after execution completes.
Test Plan: CI
Differential Revision: D73635809
Pull Request resolved: https://github.com/pytorch/pytorch/pull/157053
Approved by: https://github.com/henryoier , https://github.com/georgiaphillips
2025-06-30 00:06:37 +00:00
95a7d1912a
[sigmoid] add layout planner to executor ( #156852 )
...
Summary: if memory planning is enabled in the runtime config, we will create a copy in the executor here.
Test Plan: ci
Differential Revision: D73635622
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156852
Approved by: https://github.com/zhxchen17
2025-06-25 21:41:09 +00:00
e98dd95446
[nativert] Move SerialGraphExecutor to PyTorch core ( #156459 )
...
Summary: `SerialGraphExecutor` inherits from `GraphExecutorBase` and executes all nodes in the graph in a serial manner
Test Plan:
CI
Rollback Plan:
Differential Revision: D76917966
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156459
Approved by: https://github.com/zhxchen17 , https://github.com/jingsh
2025-06-21 01:32:06 +00:00
