OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-07 00:21:07 +01:00
Code Issues Actions 37 Packages Projects Releases Wiki Activity
66,257 Commits 8,304 Branches 150 Tags 1.5 GiB
c46fc46dba
Commit Graph

23 Commits

Author SHA1 Message Date
drisspg
c46fc46dba expose mem-eff to autograd (#110495) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110495
Approved by: https://github.com/jbschlosser
 2023-11-13 17:47:40 +00:00
Peter Bell
bbd5b935e4 Use pytree.tree_leaves everywhere (#112324) 
This changes all the instances I could find of `tree_flatten(...)[0]` or
`x, _ = tree_flatten` to use `tree_leaves`.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/112324
Approved by: https://github.com/lezcano
ghstack dependencies: #112327, #112323
 2023-10-30 03:39:04 +00:00
lezcano
c8a5bb451e Do not import sympy within torch._prims_common (#112034) 
This is the first of a few PRs that avoid importing SymPy at import time.
The pitch here is that we (almost!) do not have SymPy on our API, so
this should be feasible.

This should speed-up torch imports by a good 15% as per
https://dev-discuss.pytorch.org/t/delving-into-what-happens-when-you-import-torch/1589

In this PR we just move a few global imports into local imports.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112034
Approved by: https://github.com/ezyang
 2023-10-26 12:53:25 +00:00
cdzhan
46000bede6 Fix a typo in fake tensor test. (#111193) 
Fixes #ISSUE_NUMBER

Pull Request resolved: https://github.com/pytorch/pytorch/pull/111193
Approved by: https://github.com/janeyx99
 2023-10-17 03:36:28 +00:00
drisspg
e0dbaa04d2 Fix the meta func for mem_eff_backward (#110893) 
Fixes #110832

Pull Request resolved: https://github.com/pytorch/pytorch/pull/110893
Approved by: https://github.com/eellison
 2023-10-11 02:58:54 +00:00
Edward Z. Yang
55f956f1d2 optests improvements based on torchvision usage on nms (#108929) 
- Update cross-ref FakeMode test to use ShapeEnv.  Dynamic ops can now
  return an unbacked SymInt.  We always accept this as equal to whatever
  the real value was.
- Relax test so it works on all classes, not just unittest.TestCase
- Properly wrap the original method, so things like
  pytree.mark.parametrize are carried over
- Support dynamic shapes by default for make_fx `tracing_mode="fake"` without symbolifying everything else

Fixes https://github.com/pytorch/pytorch/issues/108927

Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/108929
Approved by: https://github.com/zou3519
 2023-09-13 13:26:15 +00:00
eellison
3495f0c999 Generate mypy hints for torch.Tag, add a couple of pointwise ops (#106910) 
Replace https://github.com/pytorch/pytorch/pull/106739, since i had a bad CLA commit.

- adds clone, and convert_element_dtype to pointwise
- adds codegen for mypy hints of torch.Tag and removes existing ignores for them

Pull Request resolved: https://github.com/pytorch/pytorch/pull/106910
Approved by: https://github.com/mlazos
 2023-08-10 05:12:27 +00:00
drisspg
f533791cd0 [SDPA] Mirror c++ implementation in FlashAttention meta func (#106477) 
# Summary
Test edge case and update meta function to match the c++ implementation
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106477
Approved by: https://github.com/eellison
 2023-08-03 00:28:27 +00:00
Richard Zou
08a054649c [operator_compile_check] Add FakeTensor testing (#103595) 
This PR adds dedicated FakeTensor testing to operator_compile_check. We
reuse CrossRefFakeMode to do this and improve the error messages on it.

Note that this only really runs detailed tests for operators that do not
have data-dependent output shape. In the future we should add something
like a dynamic CrossRefFakeMode.

Test Plan:
- existing tests (these now have improved error messages).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/103595
Approved by: https://github.com/ezyang, https://github.com/soulitzer
 2023-06-16 16:55:51 +00:00
Nikita Shulga
fd3a7264ae [MPS] Add group_norm[fwd+backward] and mean_var (take 2) (#91190) 
Use Prims to implement group_norm, group_norm_backward and mean_var

Use `torch._ops.ops` instead of `torch.ops` in numerous subpackages in
order to be able to make them importable from `torch/backend/mps/__init__.py` as this alias is defined in
15af4b1cee/torch/__init__.py (L1095)
is executed last during init process.

Add `__all__` to `torch/backends/mps/__init__.py` as well as alias all imports as private

Add `TestNNMPS.test_group_norm_backward` that validates no NaNs are generated during the backward pass

Fixes https://github.com/pytorch/pytorch/issues/88331
Pull Request resolved: https://github.com/pytorch/pytorch/pull/91190
Approved by: https://github.com/albanD
 2022-12-22 08:54:37 +00:00
PyTorch MergeBot
645eda0a00 Revert "[MPS] Add group_norm[fwd+backward] and mean_var (#91190)" 
This reverts commit 371716eb36.

Reverted https://github.com/pytorch/pytorch/pull/91190 on behalf of https://github.com/kit1980 due to Broke test_correct_module_names because of underscore _ops
 2022-12-21 19:37:43 +00:00
Nikita Shulga
371716eb36 [MPS] Add group_norm[fwd+backward] and mean_var (#91190) 
Use Prims to implement group_norm, group_norm_backward and mean_var

Use `torch._ops.ops` instead of `torch.ops` in numerous subpackages in
order to be able to make them importable from `torch/backend/mps/__init__.py` as this alias is defined in
15af4b1cee/torch/__init__.py (L1095)
is executed last during init process.

Depends on https://github.com/pytorch/pytorch/pull/91203

Fixes https://github.com/pytorch/pytorch/issues/88331
Pull Request resolved: https://github.com/pytorch/pytorch/pull/91190
Approved by: https://github.com/albanD
 2022-12-21 17:33:27 +00:00
Ram Rachum
351d73b97f Fix exception causes all over the codebase (#90271) 
This is the continuation to #90134 and hopefully the final PR in this series.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/90271
Approved by: https://github.com/kit1980
 2022-12-07 04:29:00 +00:00
Kurt Mohler
ee28b865ee Deprecate TypedStorage, its derived classes, and all of their public methods (#85303) 
Part of #85302

Pull Request resolved: https://github.com/pytorch/pytorch/pull/85303
Approved by: https://github.com/ezyang
 2022-11-08 18:11:01 +00:00
Elias Ellison
6a2b12dd65 Turn on aliasing tests for fake backwards, Fix Batch norm running mean/var decomp aliasing (#85471) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85471
Approved by: https://github.com/ezyang
 2022-09-28 23:06:59 +00:00
Elias Ellison
bcc544e9d7 Add FakeCrossRef tests for backwards, Fix Layer Norm Backward Decomp (#85417) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85417
Approved by: https://github.com/ezyang
 2022-09-26 17:08:14 +00:00
PyTorch MergeBot
d10de31cc8 Revert "Add FakeCrossRef tests for backwards, Fix Layer Norm Backward Decomp (#85417)" 
This reverts commit 78afa0cf0c.

Reverted https://github.com/pytorch/pytorch/pull/85417 on behalf of https://github.com/clee2000 due to broke tests on trunk 78afa0cf0c
 2022-09-23 17:21:43 +00:00
PyTorch MergeBot
3b195fd33e Revert "Turn on aliasing tests for fake backwards, Fix Batch norm running mean/var decomp aliasing (#85471)" 
This reverts commit 1e92eb8068.

Reverted https://github.com/pytorch/pytorch/pull/85471 on behalf of https://github.com/clee2000 due to stacked prs https://github.com/pytorch/pytorch/pull/85417 and https://github.com/pytorch/pytorch/pull/85434 broke trunk, reverting this so i can revert the others
 2022-09-23 17:13:35 +00:00
Elias Ellison
1e92eb8068 Turn on aliasing tests for fake backwards, Fix Batch norm running mean/var decomp aliasing (#85471) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85471
Approved by: https://github.com/ezyang
 2022-09-23 16:02:15 +00:00
Elias Ellison
78afa0cf0c Add FakeCrossRef tests for backwards, Fix Layer Norm Backward Decomp (#85417) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85417
Approved by: https://github.com/ezyang
 2022-09-23 15:50:03 +00:00
PyTorch MergeBot
5043457a8e Revert "Add FakeCrossRef tests for backwards, Fix Layer Norm Backward Decomp (#85417)" 
This reverts commit 9c77083965.

Reverted https://github.com/pytorch/pytorch/pull/85417 on behalf of https://github.com/clee2000 due to broke tests on trunk (and pull somehow) 9c77083965
 2022-09-22 15:44:38 +00:00
Elias Ellison
9c77083965 Add FakeCrossRef tests for backwards, Fix Layer Norm Backward Decomp (#85417) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85417
Approved by: https://github.com/ezyang
 2022-09-22 13:03:57 +00:00
Elias Ellison
d9aa6dfe88 Add Fake Cross Ref Mode, migrate sparse to it (#85382) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85382
Approved by: https://github.com/ezyang
 2022-09-21 17:15:47 +00:00