Commit Graph

5 Commits

Author SHA1 Message Date
Yuanhao Ji
16bc82a015 [Dynamo] Replace torch._dynamo.optimize() with torch.compile() [6/N] (#140688)
related commits:

- #139706
- #140238
- #140247
- #140253
- #140663
- #140688

Pull Request resolved: https://github.com/pytorch/pytorch/pull/140688
Approved by: https://github.com/williamwen42
2024-11-18 04:09:09 +00:00
Guilherme Leobas
7d283309d8 Avoid calling realize() on LazyVariableTracker on reconstruct (#138495)
Fixes: https://github.com/pytorch/pytorch/issues/137686

Pull Request resolved: https://github.com/pytorch/pytorch/pull/138495
Approved by: https://github.com/zou3519
2024-10-25 19:01:15 +00:00
Guilherme Leobas
52c917b0ba Optimize dict reconstruct to not codegen untouched values (#134876)
PR changes how `reconstruct` is done for a ConstDict. As of today, it works as follow:
(1) codegen(...) each pair of key/value
(2) create a new dictionary to hold the new items
(3) clear the original dictionary
(4) update the original dict with the one created in (2)

We do a micro optimization in the generated bytecode to:
- Only codegen the items that changed.
- Only clear the original dictionary if a key was removed.

Fixes: #133487

Pull Request resolved: https://github.com/pytorch/pytorch/pull/134876
Approved by: https://github.com/zou3519
2024-09-23 21:45:44 +00:00
PyTorch MergeBot
3b5e2689a1 Revert "Optimize dict reconstruct to not codegen untouched values (#134876)"
This reverts commit a1a57a424d.

Reverted https://github.com/pytorch/pytorch/pull/134876 on behalf of https://github.com/jeanschmidt due to new introduced test test_reconstruct.py::ReconstructTest::test_functional_call_reconstruct is breaking internally. @zou3519 may you help get those changes merged back to main? ([comment](https://github.com/pytorch/pytorch/pull/134876#issuecomment-2355697685))
2024-09-17 13:00:01 +00:00
Guilherme Leobas
a1a57a424d Optimize dict reconstruct to not codegen untouched values (#134876)
PR changes how `reconstruct` is done for a ConstDict. As of today, it works as follow:
(1) codegen(...) each pair of key/value
(2) create a new dictionary to hold the new items
(3) clear the original dictionary
(4) update the original dict with the one created in (2)

We do a micro optimization in the generated bytecode to:
- Only codegen the items that changed.
- Only clear the original dictionary if a key was removed.

Fixes: #133487

Pull Request resolved: https://github.com/pytorch/pytorch/pull/134876
Approved by: https://github.com/zou3519
2024-09-14 23:25:28 +00:00