OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity
85,638 Commits 8,304 Branches 150 Tags 1.5 GiB
2be97c7257
Commit Graph

164 Commits

Author SHA1 Message Date
Aleksei Nikiforov
2be97c7257 Update nightly s390x builds (#149337) 
This change should fix new nightly build failures for s390x.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149337
Approved by: https://github.com/malfet
 2025-03-19 23:27:14 +00:00
Tristan Rice
dea7157160 nccl: upgrade to 2.26.2 to avoid hang on ncclCommAbort (#149351) 
Fixes #149153

Yaml generated from:

```
python .github/scripts/generate_ci_workflows.py
```

Test plan:

Repro in https://gist.github.com/d4l3k/16a19b475952bc40ddd7f2febcc297b7

```
rm -rf third_party/nccl
python setup.py develop
```

Pull Request resolved: https://github.com/pytorch/pytorch/pull/149351
Approved by: https://github.com/kwen2501, https://github.com/atalman, https://github.com/malfet
 2025-03-18 05:23:18 +00:00
Aaron Gokaslan
6856d81c60 [BE]: Update CU128 cudnn to 9.8.0.87 (#148963) 
Also cu12.6 is an on old CUDNN version, we may want to upgrade it for all the performance reasons as I don't see a manywheel linux reason to stay back on the old 9.5 release. I might split that into it's own PR. This one just updates CU126 to the latest and greatest.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148963
Approved by: https://github.com/jansel, https://github.com/eqy, https://github.com/nWEIdia, https://github.com/tinglvv, https://github.com/atalman
 2025-03-13 16:59:12 +00:00
atalman
29fd875bc1 Automate stable CUDA update and linter using min Python verison (#148912) 
1. Fixes: https://github.com/pytorch/pytorch/issues/145571 . Cuda Stable is the same cuda version that is published to pypi, also used to set Metadata section in the rest of whl scripts and tag the docker releases with latest tag.
2. Updates min python version used in linter
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148912
Approved by: https://github.com/Skylion007, https://github.com/malfet
 2025-03-12 18:12:34 +00:00
Ting Lu
2a1eeaeed8 Remove 12.4 x86 builds and 12.6 sbsa builds from nightly (#148895) 
https://github.com/pytorch/pytorch/issues/145570

redo https://github.com/pytorch/pytorch/pull/148625

Pull Request resolved: https://github.com/pytorch/pytorch/pull/148895
Approved by: https://github.com/atalman
 2025-03-10 20:55:09 +00:00
Aman Karmani
a81751d8b7 [CD] Annotate linux/arm64 cuda wheels with consistent nvidia dependencies (#145021) 
This resolves issues installing torch nightly wheels into a `uv sync`-generated `.venv`

The root cause is that the x64 and arm64 cuda nightly wheels have inconsistent metadata. This can be seen comparing `generated-linux-aarch64-binary-manywheel-nightly.yml` and `generated-linux-binary-manywheel-nightly.yml`

`uv` expects consistency:

https://github.com/astral-sh/uv/issues/10693
>Frankly, it's really not ideal that they change their dependencies from wheel to wheel.
>They could still put the dependencies there with the same platform markers they're using in the other wheel though... 🤷‍♀

https://github.com/astral-sh/uv/issues/10119#issuecomment-2559898792
>I think this is something that basically has to be solved by PyTorch. The issue is that the wheels for `2.6.0.dev20241222+cu126` don't have consistent metadata, and it's a fundamental assumption of uv that the metadata for a given version _is_ consistent.

To resolve this, I modified the arm64 nightly build workflow to add two new `PYTORCH_EXTRA_INSTALL_REQUIREMENTS` entries, under `manywheel-py3_11-cuda-aarch64-build` and `manywheel-py3_12-cuda-aarch64-build`. These are based on their equivalents in the x64 workflow for the corresponding python versions.

I used the cuda 12.6 dependencies versions for the nvidia packages, to match the `DOCKER_IMAGE: pytorch/manylinuxaarch64-builder:cuda12.6-main` being used by these jobs.

(The arm64 workflow file already had several `PYTORCH_EXTRA_INSTALL_REQUIREMENTS` entries, under various cpu wheels. I'm not sure why these are there, but I left them as-is.)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/145021
Approved by: https://github.com/seemethere, https://github.com/atalman

Co-authored-by: Eli Uriegas <eliuriegas@meta.com>
Co-authored-by: Andrey Talman <atalman@fb.com>
 2025-03-10 14:39:39 +00:00
PyTorch MergeBot
99da439d10 Revert "Remove Cuda 12.4 from nightly Binaries (#148625)" 
This reverts commit 1239176fe7.

Reverted https://github.com/pytorch/pytorch/pull/148625 on behalf of https://github.com/malfet due to Broke lint ([comment](https://github.com/pytorch/pytorch/pull/148625#issuecomment-2707415005))
 2025-03-07 20:47:45 +00:00
Ting Lu
1239176fe7 Remove Cuda 12.4 from nightly Binaries (#148625) 
https://github.com/pytorch/pytorch/issues/145570

removes cuda 12.4 nightly builds

Pull Request resolved: https://github.com/pytorch/pytorch/pull/148625
Approved by: https://github.com/atalman
 2025-03-07 18:56:04 +00:00
Irem Yuksel
61c4074df7 Add Windows Arm64 Nightly Builds (#139760) 
This PR creates 3 new worklflows for Windows Arm64 target. The workflows and outputs can be reviewed at the following links:
https://github.com/pytorch/pytorch/actions/workflows/generated-windows-arm64-binary-libtorch-release-nightly.yml
https://github.com/pytorch/pytorch/actions/workflows/generated-windows-arm64-binary-libtorch-debug-nightly.yml
https://github.com/pytorch/pytorch/actions/workflows/generated-windows-arm64-binary-wheel-nightly.yml

Pull Request resolved: https://github.com/pytorch/pytorch/pull/139760
Approved by: https://github.com/malfet

Co-authored-by: Ozan Aydin <148207261+ozanMSFT@users.noreply.github.com>
Co-authored-by: Huy Do <huydhn@gmail.com>
 2025-03-07 18:53:56 +00:00
Wang, Chuanqi
3f4311d589 [CD] Upgrade xpu runtime pypi packages version and enable windows kineto again (#148319) 
Fixes https://github.com/pytorch/pytorch/issues/145155

Pull Request resolved: https://github.com/pytorch/pytorch/pull/148319
Approved by: https://github.com/xuhancn, https://github.com/atalman
 2025-03-05 18:39:55 +00:00
Ting Lu
fe100c3c5b Add libtorch nightly build for CUDA 12.8 (#146265) 
Try removing sm50 and sm60 to shrink binary size, and resolve the ld --relink error

"Architecture support for Maxwell, Pascal, and Volta is considered feature-complete and will be frozen in an upcoming release." from 12.8 release note.

Also updating the runner for cuda 12.8 test to g4dn (T4, sm75) due to the drop in sm50/60 support.

https://github.com/pytorch/pytorch/issues/145570

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146265
Approved by: https://github.com/atalman
 2025-02-21 03:04:06 +00:00
Ting Lu
e6c86952c6 Add CUDA 12.8 windows nightly build (#147037) 
https://github.com/pytorch/pytorch/issues/145570

windows AMI is deployed to prod today, prepping the windows cuda 12.8 build

Pull Request resolved: https://github.com/pytorch/pytorch/pull/147037
Approved by: https://github.com/atalman
 2025-02-19 16:59:32 +00:00
atalman
4ece056791 Nccl update to 2.25.1 for cuda 12.4-12.8 (#146073) 
Should resolve: https://github.com/pytorch/pytorch/issues/144768
We use one common nccl version for cuda builds 12.4-12.8 : ``NCCL_VERSION=v2.25.1-1``
For CUDA 11.8 we use legacy ``NCCL_VERSION=v2.21.1-1``
We use pinned version of NCCL rather then submodule.
Move nccl location from ``third_party/nccl/nccl`` to ``third_party/nccl``

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146073
Approved by: https://github.com/Skylion007, https://github.com/malfet, https://github.com/kwen2501, https://github.com/fduwjj
 2025-02-19 03:52:26 +00:00
PyTorch MergeBot
7622e29a37 Revert "Nccl update to 2.25.1 for cuda 12.4-12.8 (#146073)" 
This reverts commit eecee5863e.

Reverted https://github.com/pytorch/pytorch/pull/146073 on behalf of https://github.com/atalman due to breaks Locally building benchmarks ([comment](https://github.com/pytorch/pytorch/pull/146073#issuecomment-2667054179))
 2025-02-18 22:23:35 +00:00
atalman
eecee5863e Nccl update to 2.25.1 for cuda 12.4-12.8 (#146073) 
Should resolve: https://github.com/pytorch/pytorch/issues/144768
We use one common nccl version for cuda builds 12.4-12.8 : ``NCCL_VERSION=v2.25.1-1``
For CUDA 11.8 we use legacy ``NCCL_VERSION=v2.21.1-1``
We use pinned version of NCCL rather then submodule.
Move nccl location from ``third_party/nccl/nccl`` to ``third_party/nccl``

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146073
Approved by: https://github.com/Skylion007, https://github.com/malfet, https://github.com/kwen2501, https://github.com/fduwjj
 2025-02-14 21:23:19 +00:00
PyTorch MergeBot
e06ee4aa9f Revert "Nccl update to 2.25.1 for cuda 12.4-12.8 (#146073)" 
This reverts commit 06f4a5c0e5.

Reverted https://github.com/pytorch/pytorch/pull/146073 on behalf of https://github.com/atalman due to breaks macos builds: ModuleNotFoundError: No module named 'torch._C._distributed_c10d'; 'torch._C' is not a package ([comment](https://github.com/pytorch/pytorch/pull/146073#issuecomment-2659802389))
 2025-02-14 16:44:46 +00:00
atalman
06f4a5c0e5 Nccl update to 2.25.1 for cuda 12.4-12.8 (#146073) 
Should resolve: https://github.com/pytorch/pytorch/issues/144768
We use one common nccl version for cuda builds 12.4-12.8 : ``NCCL_VERSION=v2.25.1-1``
For CUDA 11.8 we use legacy ``NCCL_VERSION=v2.21.1-1``
We use pinned version of NCCL rather then submodule.
Move nccl location from ``third_party/nccl/nccl`` to ``third_party/nccl``

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146073
Approved by: https://github.com/Skylion007, https://github.com/malfet, https://github.com/kwen2501, https://github.com/fduwjj
 2025-02-14 15:29:59 +00:00
Ting Lu
0344bf8a5a [cuDNN] cuDNN to 9.7.1.26 for CUDA 12.8 (#146957) 
rebasing for https://github.com/pytorch/pytorch/pull/146717

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146957
Approved by: https://github.com/Skylion007, https://github.com/eqy, https://github.com/nWEIdia, https://github.com/atalman

Co-authored-by: Aaron Gokaslan <aaronGokaslan@gmail.com>
 2025-02-12 23:43:34 +00:00
Wang, Chuanqi
8c80c13b34 [CD] Add python 3.13t build for xpu (#146614) 
Fixes #146451

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146614
Approved by: https://github.com/atalman
 2025-02-12 07:01:36 +00:00
Mikayla Gawarecki
861bf892fb Set USE_CUFILE=1 by default and add pypi package to binary build matrix (#145748) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/145748
Approved by: https://github.com/atalman
 2025-02-11 15:49:01 +00:00
Ting Lu
9e45bc82e9 [aarch64] CUDA 12.8 aarch64 builds to nightly binaries (#146378) 
https://github.com/pytorch/pytorch/issues/145570

Adding Cuda 12.8 and keeping 12.6 for the sbsa build, supported CUDA_ARCH: 9.0, 10.0, 12.0

Refactor the binaries matrix for cuda sbsa build. Previously cuda-aarch64 was hardcoded to cuda 12.6. Now reads 12.6 and 12.8, new build naming example [manywheel-py3_9-cuda-aarch64-12_8-build](https://github.com/pytorch/pytorch/actions/runs/13132625006/job/36640885079?pr=146378#logs)

TODO: once 12.8 is stable, remove 12.6 in sbsa

Pull Request resolved: https://github.com/pytorch/pytorch/pull/146378
Approved by: https://github.com/atalman
 2025-02-05 02:55:21 +00:00
atalman
a7cc6d3e84 Manylinux 2.28 migration - remove pre-cxx11 abi libtorch builds (#146200) 
Related to: https://github.com/pytorch/pytorch/issues/123649
Removing pre-cxx11 abi builds.
As per announcement : https://dev-discuss.pytorch.org/t/pytorch-linux-wheels-switching-to-new-wheel-build-platform-manylinux-2-28-on-november-12-2024/2581
Pull Request resolved: https://github.com/pytorch/pytorch/pull/146200
Approved by: https://github.com/kit1980, https://github.com/huydhn
 2025-01-31 21:43:12 +00:00
Ting Lu
9232355bb0 Add CUDA 12.8 manywheel x86 Builds to Binaries Matrix (#145792) 
https://github.com/pytorch/pytorch/issues/145570

Adding cuda 12.8.0 x86 builds first
Pull Request resolved: https://github.com/pytorch/pytorch/pull/145792
Approved by: https://github.com/nWEIdia, https://github.com/malfet, https://github.com/atalman
 2025-01-31 16:12:02 +00:00
Aaron Orenstein
60f98262f1 PEP585: .github (#145707) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/145707
Approved by: https://github.com/huydhn
 2025-01-27 21:21:01 +00:00
atalman
2cffbff7da Add 3.13t Windows and MacOS binary builds (#141806) 
Related to: https://github.com/pytorch/pytorch/issues/130249

For conda uses approach described here:
https://conda-forge.org/blog/2024/09/26/python-313/

Create Python 3.13t conda env like so:
```
conda create -n py313 python=3.13 python-freethreading  -c conda-forge
```

For windows executable installation we need to pass additional parameter to enable 3.13t:
```
Include_freethreaded=1
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141806
Approved by: https://github.com/albanD
 2025-01-21 17:16:19 +00:00
Wang, Chuanqi
b4b4e57469 [CD] Enable profiling for XPU Windows nightly wheels (#144316) 
PR https://github.com/pytorch/pytorch/pull/144034 added profiling support for torch XPU Windows binary, enable it in PyTorch XPU Windows CD
Works for https://github.com/pytorch/pytorch/issues/114850

Pull Request resolved: https://github.com/pytorch/pytorch/pull/144316
Approved by: https://github.com/xuhancn, https://github.com/atalman
 2025-01-14 19:01:27 +00:00
Nikita Shulga
6053242890 [CD] Enable python3.13t builds for aarch64 (#144698) 
But make sure that right numpy version is picked (2.0.2 does not support 3.13)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/144698
Approved by: https://github.com/atalman
ghstack dependencies: #144696, #144697, #144716
 2025-01-14 02:29:01 +00:00
Nikita Shulga
d44c3906b8 [EZ] [CD] Add 3.13 to FULL_PYTHON_VERSIONS (#144697) 
Separation was necessary for Conda codegen, but now it's gone
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144697
Approved by: https://github.com/atalman, https://github.com/izaitsevfb
ghstack dependencies: #144696
 2025-01-13 19:12:12 +00:00
Nikita Shulga
d2f905760d [EZ] [CD] Eliminate stale TODO (#144696) 
As 3.13 has been enabled across the board, which one can verify by running `./github/regenerate.sh` and observe that non of the configs have changed

Pull Request resolved: https://github.com/pytorch/pytorch/pull/144696
Approved by: https://github.com/izaitsevfb, https://github.com/atalman
 2025-01-13 19:12:12 +00:00
Xuehai Pan
b6bdb67f82 [BE][Easy] use pathlib.Path instead of dirname / ".." / pardir (#129374) 
Changes by apply order:

1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.

    `.parent{...}.absolute()` -> `.absolute().parent{...}`

4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)

    `.parent.parent.parent.parent` -> `.parents[3]`

5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~

    ~`.parents[3]` -> `.parents[4 - 1]`~

6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~

Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
 2024-12-29 17:23:13 +00:00
Xuehai Pan
c4bff71854 [Easy] Add ROCm support to nightly pull tool (#141282) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141282
Approved by: https://github.com/malfet
ghstack dependencies: #143263
 2024-12-27 00:07:38 +00:00
PyTorch MergeBot
475656fd9c Revert "[BE][Easy] use pathlib.Path instead of dirname / ".." / pardir (#129374)" 
This reverts commit 2293fe1024.

Reverted https://github.com/pytorch/pytorch/pull/129374 on behalf of https://github.com/malfet due to failing internal ROCM builds with error: ModuleNotFoundError: No module named hipify ([comment](https://github.com/pytorch/pytorch/pull/129374#issuecomment-2562973920))
 2024-12-26 17:32:23 +00:00
Jithun Nair
75e1f8a227 [ROCm] upgrade nightly wheels to rocm6.3 - 2 of 2 (binaries) (#143613) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/143613
Approved by: https://github.com/jeffdaily
 2024-12-23 19:47:30 +00:00
Xuehai Pan
2293fe1024 [BE][Easy] use pathlib.Path instead of dirname / ".." / pardir (#129374) 
Changes by apply order:

1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.

    `.parent{...}.absolute()` -> `.absolute().parent{...}`

4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)

    `.parent.parent.parent.parent` -> `.parents[3]`

5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~

    ~`.parents[3]` -> `.parents[4 - 1]`~

6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~

Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
 2024-12-21 22:08:01 +00:00
atalman
2400db115c Use Manylinux 2.28 for nightly build and cxx11-abi (#143423) 
As per: https://dev-discuss.pytorch.org/t/pytorch-linux-wheels-switching-to-new-wheel-build-platform-manylinux-2-28-on-november-12-2024/2581

Linux Builds: CPU, CUDA 11.8, CUDA 12.4 switched to Manylinux 2.28 and D_GLIBCXX_USE_CXX11_ABI=1 on the week of Dec 16

Pull Request resolved: https://github.com/pytorch/pytorch/pull/143423
Approved by: https://github.com/huydhn, https://github.com/malfet, https://github.com/seemethere
 2024-12-18 02:02:58 +00:00
Ting Lu
f26b75b7ac [aarch64] add CUDA 12.6 sbsa nightly binary (#142335) 
related to #138440

Pull Request resolved: https://github.com/pytorch/pytorch/pull/142335
Approved by: https://github.com/atalman
 2024-12-10 06:19:28 +00:00
Jithun Nair
a1b5067297 Enable py3.13 wheels for ROCm (#142294) 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/142294
Approved by: https://github.com/huydhn
 2024-12-10 01:10:24 +00:00
chuanqiw
b64a537993 [CD] xpu nightly manylinux whl with cxx11-abi (#142210) 
Follow https://github.com/pytorch/pytorch/issues/123649
Works for https://github.com/pytorch/pytorch/issues/114850
Pull Request resolved: https://github.com/pytorch/pytorch/pull/142210
Approved by: https://github.com/EikanWang, https://github.com/atalman, https://github.com/malfet
 2024-12-06 15:10:47 +00:00
atalman
c6c45467a3 Use cxx11-abi for Linux CUDA 12.6 builds (#142064) 
Manylinux 2.28 and cxx11-abi migration. Please see: https://dev-discuss.pytorch.org/t/pytorch-linux-wheels-switching-to-new-wheel-build-platform-manylinux-2-28-on-november-12-2024/2581
Pull Request resolved: https://github.com/pytorch/pytorch/pull/142064
Approved by: https://github.com/kit1980, https://github.com/malfet
 2024-12-05 14:51:50 +00:00
Jithun Nair
9dffd12f90 Upgrade ROCm wheels to manylinux2_28 - 2 of 2 (binaries) (#141423) 
Depends on https://github.com/pytorch/pytorch/pull/140681 and https://github.com/pytorch/pytorch/pull/141609

Highlights:
* Upgrade binaries to ROCm6.2.4 to use latest docker images
* Remove pre-cxx11 builds for libtorch on ROCm
* Use manylinux2_28 docker images for ROCm
* Set `DESIRED_DEVTOOLSET=cxx-abi` (and hence `_GLIBCXX_USE_CXX11_ABI=1`) for ROCm manylinux2_28 wheels (ROCm RHEL8 packages also have GCC_ABI=1, so it keeps it consistent)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/141423
Approved by: https://github.com/jeffdaily

Co-authored-by: Jeff Daily <jeff.daily@amd.com>
Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com>
 2024-12-04 07:00:25 +00:00
Ting Lu
e5f5283ab2 Fix cuda arch full version for 12.6 (#141976) 
follow up for https://github.com/pytorch/pytorch/pull/141433/files
build still showing up as 12.6.2 in the name, see latest https://github.com/pytorch/pytorch/actions/runs/12134985224/job/33833276884.

related to https://github.com/pytorch/pytorch/issues/138440

Pull Request resolved: https://github.com/pytorch/pytorch/pull/141976
Approved by: https://github.com/atalman, https://github.com/nWEIdia, https://github.com/Skylion007
 2024-12-03 20:33:01 +00:00
atalman
0f3f801fc2 Add windows CUDA 12.6 nightly builds (#141805) 
Windows AMI was published to prod. This PR adds CUDA 12.6 nightly builds

Pull Request resolved: https://github.com/pytorch/pytorch/pull/141805
Approved by: https://github.com/huydhn, https://github.com/Skylion007
 2024-11-30 14:39:47 +00:00
chuanqiw
a23ac6f8bd [CD] Enable pypi dependencies both for XPU linux and Windows whls (#141135) 
Enable xpu runtime pypi packages as dependencies of XPU CD wheels both for Linux and Windows.
Fixes https://github.com/pytorch/pytorch/issues/135867
Works for https://github.com/pytorch/pytorch/issues/139722 and https://github.com/pytorch/pytorch/issues/114850
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141135
Approved by: https://github.com/atalman
 2024-11-29 21:35:07 +00:00
Aaron Gokaslan
7224cd4471 [BE]: Update 12.6 builds to CUDA 12.6.3 (#141433) 
Update CUDA 12.6 to Update 3 and make cusparse-lt 0.6.3? #141365 Was going to leave some comments on #141365, but though it was just faster to open a PR here.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141433
Approved by: https://github.com/atalman
 2024-11-28 22:01:47 +00:00
atalman
893a4390c9 Use cuda 12.6 wheels with Manylinux 2.28. Use Manylinux2014 for CPU, CUDA11.8, CUDA12.4 (#141565) 
For release 2.6 we will be using only CUDA 12.6 binaries on Manylinux 2.28.
Issue: https://github.com/pytorch/pytorch/issues/123649
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141565
Approved by: https://github.com/Skylion007, https://github.com/huydhn, https://github.com/malfet
 2024-11-26 19:36:42 +00:00
Xuehai Pan
2a6eaa2e6f Refactor nightly pull tool to use venv and pip (#141281) 
Resolves #141238

- #141238

Example output:

```console
$ python3.12 tools/nightly.py checkout -b my-nightly-branch -p my-env --python python3.10
log file: /Users/PanXuehai/Projects/pytorch/nightly/log/2024-11-22_04h15m45s_63f8b29e-a845-11ef-bbf9-32c784498a7b/nightly.log
Creating virtual environment
Creating venv (Python 3.10.15): /Users/PanXuehai/Projects/pytorch/my-env
Installing packages
Upgrading package(s) (https://download.pytorch.org/whl/nightly/cpu): pip, setuptools, wheel
Installing packages took 5.576 [s]
Creating virtual environment took 9.505 [s]
Downloading packages
Downloading package(s) (https://download.pytorch.org/whl/nightly/cpu): torch
Downloaded 9 file(s) to /var/folders/sq/7sf73d5s2qnb3w6jjsmhsw3h0000gn/T/pip-download-lty5dvz4:
  - mpmath-1.3.0-py3-none-any.whl
  - torch-2.6.0.dev20241121-cp310-none-macosx_11_0_arm64.whl
  - jinja2-3.1.4-py3-none-any.whl
  - sympy-1.13.1-py3-none-any.whl
  - MarkupSafe-3.0.2-cp310-cp310-macosx_11_0_arm64.whl
  - networkx-3.4.2-py3-none-any.whl
  - fsspec-2024.10.0-py3-none-any.whl
  - filelock-3.16.1-py3-none-any.whl
  - typing_extensions-4.12.2-py3-none-any.whl
Downloading packages took 7.628 [s]
Installing dependencies
Installing packages
Installing package(s) (https://download.pytorch.org/whl/nightly/cpu): numpy, cmake, ninja, packaging, ruff, mypy, pytest, hypothesis, ipython, rich, clang-format, clang-tidy, sphinx, mpmath-1.3.0-py3-none-any.whl, jinja2-3.1.4-py3-none-any.whl, sympy-1.13.1-py3-none-any.whl, MarkupSafe-3.0.2-cp310-cp310-macosx_11_0_arm64.whl, networkx-3.4.2-py3-none-any.whl, fsspec-2024.10.0-py3-none-any.whl, filelock-3.16.1-py3-none-any.whl, typing_extensions-4.12.2-py3-none-any.whl
Installing packages took 42.514 [s]
Installing dependencies took 42.515 [s]
Unpacking wheel file
Unpacking wheel file took 3.223 [s]
Checking out nightly PyTorch
Found released git version ac47a2d9714278889923ddd40e4210d242d8d4ee
Found nightly release version e0482fdf95
Switched to a new branch 'my-nightly-branch'
Checking out nightly PyTorch took 0.198 [s]
Moving nightly files into repo
Linking /var/folders/sq/7sf73d5s2qnb3w6jjsmhsw3h0000gn/T/wheel-dljxil5i/torch-2.6.0.dev20241121/torch/_C.cpython-310-darwin.so -> /Users/PanXuehai/Projects/pytorch/torch/_C.cpython-310-darwin.so
Linking /var/folders/sq/7sf73d5s2qnb3w6jjsmhsw3h0000gn/T/wheel-dljxil5i/torch-2.6.0.dev20241121/torch/lib/libtorch_python.dylib -> /Users/PanXuehai/Projects/pytorch/torch/lib/libtorch_python.dylib
...
Linking /var/folders/sq/7sf73d5s2qnb3w6jjsmhsw3h0000gn/T/wheel-dljxil5i/torch-2.6.0.dev20241121/torch/include/c10/macros/Macros.h -> /Users/PanXuehai/Projects/pytorch/torch/include/c10/macros/Macros.h
Moving nightly files into repo took 11.426 [s]
Writing pytorch-nightly.pth
Writing pytorch-nightly.pth took 0.036 [s]
-------
PyTorch Development Environment set up!
Please activate to enable this environment:

  $ source /Users/PanXuehai/Projects/pytorch/my-env/bin/activate
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141281
Approved by: https://github.com/seemethere
 2024-11-22 20:03:55 +00:00
Aaron Gokaslan
765a347d21 [BE]: Update CUDNN for Linux to 9.5.1.17 for 12.6 only (#137978) 
* Significantly faster, better CUDNN Attention especially on Hopper (FA3 implementation?)
* Lots of bugfixes
* Better performance
* More numerically stable / fixed heuristics
* More functionality for SDPA

Pull Request resolved: https://github.com/pytorch/pytorch/pull/137978
Approved by: https://github.com/eqy, https://github.com/drisspg, https://github.com/nWEIdia, https://github.com/atalman, https://github.com/malfet
 2024-11-20 23:11:39 +00:00
atalman
99a03211cb Deprecate conda nightly builds (#141024) 
Removing CD as per https://github.com/pytorch/pytorch/issues/138506

Pull Request resolved: https://github.com/pytorch/pytorch/pull/141024
Approved by: https://github.com/malfet
 2024-11-19 16:09:54 +00:00
atalman
cec82c3aed Use Manylinux 2.28 for aarch64 CPU workflows (#140743) 
Use https://hub.docker.com/r/pytorch/manylinux2_28_aarch64-builder/tags

Similar to https://github.com/pytorch/pytorch/pull/138732
Pull Request resolved: https://github.com/pytorch/pytorch/pull/140743
Approved by: https://github.com/malfet
 2024-11-15 01:46:29 +00:00
atalman
70acf02116 Use Manylinux2_28 for wheel builds (#138732) 
Fixes https://github.com/pytorch/pytorch/issues/123649
Use Manylinux 2_28 Docker builds for PyTorch Nightly builds

This moves the wheels to a Docker image that uses : ``quay.io/pypa/manylinux_2_28_x86_64`` as a base rather then ``centos:7`` which is EOL on June 30, 2024.

Information:
https://github.com/pypa/manylinux#manylinux_2_28-almalinux-8-based

manylinux_2_28 (AlmaLinux 8 based)
Toolchain: GCC 13
Built wheels are also expected to be compatible with other distros using glibc 2.28 or later, including:
Debian 10+
Ubuntu 18.10+
Fedora 29+
CentOS/RHEL 8+

This migration should enable us to migrate to latest CUDNN version, and land this PR: https://github.com/pytorch/pytorch/pull/137978

Pull Request resolved: https://github.com/pytorch/pytorch/pull/138732
Approved by: https://github.com/Skylion007, https://github.com/malfet, https://github.com/huydhn
 2024-11-14 00:25:47 +00:00