frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-21 05:34:18 +08:00
Code Issues Packages Projects Releases Wiki Activity
94,736 Commits 5,442 Branches 1,295 Tags
main
Commit Graph

11 Commits

Author SHA1 Message Date
Sherlock Huang
e532f62e0d Introduce joint_custom_pass callback (#164981) 
```
        def joint_custom_pass(joint_gm: torch.fx.GraphModule, joint_inputs):
           # apply your pass for joint graph here

            return joint_gm

        class M(torch.nn.Module):
            def forward(self, x):
                return x.sin()

        x = torch.randn(10, requires_grad=False)
        compiled_fn = torch.compile(M(), backend="aot_eager")

        with torch._functorch.config.patch("joint_custom_pass", joint_custom_pass):
            out = compiled_fn(x)
```

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164981
Approved by: https://github.com/ezyang, https://github.com/anijain2305
 2025-10-09 04:40:54 +00:00
Brian Hirsh
a2b0b2698d inductor codecache: include private inductor configs in cache key (#153672) 
Fixes https://github.com/pytorch/torchtitan/issues/1185

It looks like inductor's logic to include inductor configs in the cache key skips configs with a leading underscore by default. This came up in torchtitan - there's an asyncTP pipelining pass in inductor gated by a private config, and by not caching on the config we were attempting to use asyncTP when we shouldn't be.

I'm not sure how worried we should be on the blast radius of this change. On the one hand:

(1) it technically fixes any silent correctness issues in the cache around any other private inductor configs (it looks like there are a few)

(2) there is some risk that there are some "harmless" configs that we are now including in the key, which may increase false negatives. I do see that there is an explicit list for "configs we want to ignore for caching" (`_save_config_ignore`), so my hope is that all harmless configs are already encapsulated there.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/153672
Approved by: https://github.com/oulgen
 2025-06-11 01:33:24 +00:00
PyTorch MergeBot
31f95b5d2e Revert "inductor codecache: include private inductor configs in cache key (#153672)" 
This reverts commit 2c1cb38d9516e10474b4f12a2e839046648a71a8.

Reverted https://github.com/pytorch/pytorch/pull/153672 on behalf of https://github.com/malfet due to Looks like it regressed pr_time_benchmarks, see ba3f91af97/1 ([comment](https://github.com/pytorch/pytorch/pull/153672#issuecomment-2922759739))
 2025-05-30 15:54:14 +00:00
Brian Hirsh
2c1cb38d95 inductor codecache: include private inductor configs in cache key (#153672) 
Fixes https://github.com/pytorch/torchtitan/issues/1185

It looks like inductor's logic to include inductor configs in the cache key skips configs with a leading underscore by default. This came up in torchtitan - there's an asyncTP pipelining pass in inductor gated by a private config, and by not caching on the config we were attempting to use asyncTP when we shouldn't be.

I'm not sure how worried we should be on the blast radius of this change. On the one hand:

(1) it technically fixes any silent correctness issues in the cache around any other private inductor configs (it looks like there are a few)

(2) there is some risk that there are some "harmless" configs that we are now including in the key, which may increase false negatives. I do see that there is an explicit list for "configs we want to ignore for caching" (`_save_config_ignore`), so my hope is that all harmless configs are already encapsulated there.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/153672
Approved by: https://github.com/oulgen
ghstack dependencies: #153766
 2025-05-30 00:24:29 +00:00
Xuehai Pan
56935684c3 Use Generic TypeAlias (PEP 585) and Union Type (PEP 604) in .pyi stub files (#129419) 
------

- [Generic TypeAlias (PEP 585)](https://peps.python.org/pep-0585): e.g. `typing.List[T] -> list[T]`, `typing.Dict[KT, VT] -> dict[KT, VT]`, `typing.Type[T] -> type[T]`.
- [Union Type (PEP 604)](https://peps.python.org/pep-0604): e.g. `Union[X, Y] -> X | Y`, `Optional[X] -> X | None`, `Optional[Union[X, Y]] -> X | Y | None`.

Note that in `.pyi` stub files, we do not need `from __future__ import annotations`. So this PR does not violate issue #117449:

- #117449

Pull Request resolved: https://github.com/pytorch/pytorch/pull/129419
Approved by: https://github.com/ezyang
ghstack dependencies: #129375, #129376
 2024-06-29 09:23:39 +00:00
PyTorch MergeBot
83caf4960f Revert "Use Generic TypeAlias (PEP 585) and Union Type (PEP 604) in .pyi stub files (#129419)" 
This reverts commit e40f50cb87bcd176a380b729af5dda13dbe9c399.

Reverted https://github.com/pytorch/pytorch/pull/129419 on behalf of https://github.com/huydhn due to Sorry for reverting your change but I need to revert to cleanly revert https://github.com/pytorch/pytorch/pull/129374, please do a rebase and reland this ([comment](https://github.com/pytorch/pytorch/pull/129375#issuecomment-2197800541))
 2024-06-29 00:44:24 +00:00
Xuehai Pan
e40f50cb87 Use Generic TypeAlias (PEP 585) and Union Type (PEP 604) in .pyi stub files (#129419) 
------

- [Generic TypeAlias (PEP 585)](https://peps.python.org/pep-0585): e.g. `typing.List[T] -> list[T]`, `typing.Dict[KT, VT] -> dict[KT, VT]`, `typing.Type[T] -> type[T]`.
- [Union Type (PEP 604)](https://peps.python.org/pep-0604): e.g. `Union[X, Y] -> X | Y`, `Optional[X] -> X | None`, `Optional[Union[X, Y]] -> X | Y | None`.

Note that in `.pyi` stub files, we do not need `from __future__ import annotations`. So this PR does not violate issue #117449:

- #117449

Pull Request resolved: https://github.com/pytorch/pytorch/pull/129419
Approved by: https://github.com/ezyang
ghstack dependencies: #129375, #129376
 2024-06-28 15:37:57 +00:00
Aaron Orenstein
8db9dfa2d7 Flip default value for mypy disallow_untyped_defs [9/11] (#127846) 
See #127836 for details.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/127846
Approved by: https://github.com/ezyang
ghstack dependencies: #127842, #127843, #127844, #127845
 2024-06-08 18:50:06 +00:00
Oguz Ulgen
52bcf120e5 Make inductor config hashing more portable (#127022) 
Summary: masnesral and I noticed that config contains non portable artifacts. Lets fix that.

Test Plan: adhoc testing

Differential Revision: D57748025

Pull Request resolved: https://github.com/pytorch/pytorch/pull/127022
Approved by: https://github.com/masnesral
 2024-05-25 03:01:33 +00:00
David Berard
5c0976fa04 Revert "[dynamo] guarded config (#111299)" (#115386) 
This reverts commit 5927e9cbf2ac18aaaaecaab02258b7a35ac10969.

Differential Revision: [D51959266](https://our.internmc.facebook.com/intern/diff/D51959266)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/115386
Approved by: https://github.com/yanboliang, https://github.com/malfet
ghstack dependencies: #115384, #115401, #115385
 2023-12-11 19:35:42 +00:00
Jon Chuang
00b67193ef [utils] move config_typing.pyi to torch.utils (#113929) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/113929
Approved by: https://github.com/ezyang, https://github.com/jansel
ghstack dependencies: #111299, #111300, #113901, #113916
 2023-11-17 18:51:57 +00:00