Skip to content

Add XPU to binary build generation#5489

Merged
atalman merged 4 commits intopytorch:mainfrom
chuanqi129:xpu_binary
Jul 23, 2024
Merged

Add XPU to binary build generation#5489
atalman merged 4 commits intopytorch:mainfrom
chuanqi129:xpu_binary

Conversation

@chuanqi129
Copy link
Copy Markdown
Contributor

@chuanqi129 chuanqi129 commented Jul 22, 2024 •
edited
Loading

Torch XPU nightly wheel build has been enabled with #pytorch/pytorch#129560 landed, add XPU to binary build generation. Works for pytorch/pytorch#114850

> python tools/scripts/generate_binary_build_matrix.py --with-xpu enable
{
  "include": [
    {
      "python_version": "3.8",
      "gpu_arch_type": "cpu",
      "gpu_arch_version": "",
      "desired_cuda": "cpu",
      "container_image": "pytorch/manylinux-builder:cpu",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-cpu",
      "validation_runner": "linux.2xlarge",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "cuda",
      "gpu_arch_version": "11.8",
      "desired_cuda": "cu118",
      "container_image": "pytorch/manylinux-builder:cuda11.8",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-cuda11_8",
      "validation_runner": "linux.g5.4xlarge.nvidia.gpu",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu118",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "cuda",
      "gpu_arch_version": "12.1",
      "desired_cuda": "cu121",
      "container_image": "pytorch/manylinux-builder:cuda12.1",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-cuda12_1",
      "validation_runner": "linux.g5.4xlarge.nvidia.gpu",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "cuda",
      "gpu_arch_version": "12.4",
      "desired_cuda": "cu124",
      "container_image": "pytorch/manylinux-builder:cuda12.4",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-cuda12_4",
      "validation_runner": "linux.g5.4xlarge.nvidia.gpu",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu124",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "rocm",
      "gpu_arch_version": "6.0",
      "desired_cuda": "rocm6.0",
      "container_image": "pytorch/manylinux-builder:rocm6.0",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-rocm6_0",
      "validation_runner": "linux.2xlarge",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.0",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "rocm",
      "gpu_arch_version": "6.1",
      "desired_cuda": "rocm6.1",
      "container_image": "pytorch/manylinux-builder:rocm6.1",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-rocm6_1",
      "validation_runner": "linux.2xlarge",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.1",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    },
    {
      "python_version": "3.8",
      "gpu_arch_type": "xpu",
      "gpu_arch_version": "",
      "desired_cuda": "xpu",
      "container_image": "pytorch/manylinux2_28-builder:xpu",
      "package_type": "manywheel",
      "build_name": "manywheel-py3_8-xpu",
      "validation_runner": "linux.2xlarge",
      "installation": "pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu",
      "channel": "nightly",
      "upload_to_base_bucket": "no",
      "stable_version": "2.3.1",
      "use_split_build": false
    }
  ]
}

@vercel
Copy link
Copy Markdown

vercel bot commented Jul 22, 2024

@chuanqi129 is attempting to deploy a commit to the Meta Open Source Team on Vercel.

A member of the Team first needs to authorize it.

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jul 22, 2024
atalman
atalman reviewed Jul 22, 2024
View reviewed changes
.github/workflows/generate_binary_build_matrix.yml Outdated
type: string
with-xpu:
description: "Build with XPU?"
default: "enable"
Copy link
Copy Markdown
Contributor

@atalman atalman Jul 22, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please keep it disable for now. Since this will automatically enable vision and audio builds. I don't think we want this at this point.

Copy link
Copy Markdown
Contributor Author

@chuanqi129 chuanqi129 Jul 22, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi @atalman thanks for the review, we plan to enable vision and audio build also for now. So that we can have completed nightly install cmd like pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu. How do you think? Any concern in here?

Copy link
Copy Markdown
Contributor

@atalman atalman Jul 23, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

hi @chuanqi129 please set default to disable since multiple domains use this script. Here are some examples:
https://github.com/pytorch/data/blob/main/.github/workflows/build_wheels_linux.yml#L29
https://github.com/pytorch/torchtune/blob/main/.github/workflows/build_linux_wheels.yaml#L29
https://github.com/pytorch/ao/blob/main/.github/workflows/build_wheels_linux.yml#L30

Copy link
Copy Markdown
Contributor Author

@chuanqi129 chuanqi129 Jul 23, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh, thanks @atalman for the reminder. Got your point now, I will set the with_xpu disable as default and enable it in torchvision and torchaudio after this PR land, does it make sense to you?

Copy link
Copy Markdown
Contributor Author

@chuanqi129 chuanqi129 Jul 23, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

changed

@chuanqi129 chuanqi129 requested a review from atalman July 22, 2024 23:50
atalman
atalman approved these changes Jul 23, 2024
View reviewed changes
Copy link
Copy Markdown
Contributor

@atalman atalman left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm, please adress comment of setting with with-xpu to disable by default

@atalman atalman merged commit 7880593 into pytorch:main Jul 23, 2024
This was referenced Jul 23, 2024
Merged
Merged
@atalman atalman mentioned this pull request Jul 30, 2024
Merged
atalman added a commit that referenced this pull request Jul 30, 2024
@atalman
Add xpu test to update_tests_assets.sh (#5524)
57effcb 
Something that was missed by
#5489
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@atalman atalman atalman approved these changes

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@chuanqi129 @atalman @facebook-github-bot