8000 Release 2.5.1 validations checklist and cherry-picks · Issue #138876 · pytorch/pytorch · GitHub
[go: up one dir, main page]
Skip to content

Release 2.5.1 validations checklist and cherry-picks #138876

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
8 tasks done
kit1980 opened this issue Oct 25, 2024 · 6 comments
Closed
8 tasks done

Release 2.5.1 validations checklist and cherry-picks #138876

kit1980 opened this issue Oct 25, 2024 · 6 comments
Labels
oncall: releng In support of CI and Release Engineering
Milestone
2.5.1

Comments

@kit1980
Copy link
Contributor
kit1980 commented Oct 25, 2024 •
edited
Loading

Similar to #137492

Manual validations:

  • Python 3.13 wheel validate @kit1980

      pip3 install torch==2.5.1 --index-url https://download.pytorch.org/whl/test/cu124
  
  Successfully installed MarkupSafe-3.0.2 filelock-3.13.1 fsspec-2024.6.1 jinja2-3.1.4 mpmath-1.3.0 networkx-3.3 nvidia-cublas-cu12-12.4.5.8 nvidia-cuda-cupti-cu12-12.4.127 nvidia-cuda-nvrtc-cu12-12.4.127 nvidia-cuda-runtime-cu12-12.4.127 nvidia-cudnn-cu12-9.1.0.70 nvidia-cufft-cu12-11.2.1.3 nvidia-curand-cu12-10.3.5.147 nvidia-cusolver-cu12-11.6.1.9 nvidia-cusparse-cu12-12.3.1.170 nvidia-nccl-cu12-2.21.5 nvidia-nvjitlink-cu12-12.4.127 nvidia-nvtx-cu12-12.4.127 setuptools-70.0.0 sympy-1.13.1 torch-2.5.1+cu124 typing-extensions-4.12.2
  
  python3 -c "import torch; print(torch.__version__, torch.cuda.is_available())"
  
  /home/sdym/miniconda3/envs/py313/lib/python3.13/site-packages/torch/_subclasses/functional_tensor.py:295: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:84.)
  cpu = _conversion_method_template(device=torch.device("cpu"))
  2.5.1+cu124 True

  • CUDA pypi binaries with slimmed dependencies are usable in standard AWS containers amazonlinux2023 @kit1980

      docker run -it --gpus=all --net=host amazonlinux:2023 bash
            
  dnf update
  dnf -y install python3-pip
  pip3 install torch==2.5.1 --index-url https://download.pytorch.org/whl/test/cu124
  python3 -c "import torch; print(torch.cuda.is_available())"

  • CUDA pypi binaries with slimmed dependencies are usable on almalinux/9-base @kit1980

      docker run -it --gpus=all --net=host almalinux/9-base bash
            
  dnf update
  dnf -y install python3-pip
  pip3 install torch==2.5.1 --index-url https://download.pytorch.org/whl/test/cu124
  python3 -c "import torch; print(torch.cuda.is_available())"

  • CUDA pypi binaries with slimmed dependencies are usable on default latest Ubuntu @kit1980

      docker run -it --gpus=all --net=host ubuntu bash

  apt-get update
  apt-get install python3-pip
  pip3 install torch==2.5.1 --index-url https://download.pytorch.org/whl/test/cu124 --break-system-packages
  python3 -c "import torch; print(torch.__version__, torch.cuda.is_available())"

  • PyTorch can be imported without a warning on aarch64 system @malfet

% docker run --rm -it python:3.11 bash -c "pip install numpy torch --quiet --index-url https://download.pytorch.org/whl/test/cpu;python -c 'import torch;print(torch.__version__)'"
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv
2.5.1
  • PyTorch MPS regression crash @malfet
% python -c "import torch; x=torch.rand(16, 16, device='mps', dtype=torch.float16); print(torch.__version__, torch.backends.mps.is_macos_or_newer(15, 0)); print(x[:,0:2].view(torch.float32) + 1)"
2.5.1 True
tensor([[1.0000],
        [1.0006],
        [1.0008],
        [1.0001],
        [1.0000],
        [1.0011],
        [1.0000],
        [1.0000],
        [1.0000],
        [1.0016],
        [1.0000],
        [1.0000],
        [1.0001],
        [1.0001],
        [1.0000],
        [1.0004]], device='mps:0')
@malfet malfet added the oncall: releng In support of CI and Release Engineering label Oct 25, 2024
@kit1980 kit1980 added this to the 2.5.1 milestone Oct 25, 2024
@Skylion007
Copy link
Collaborator

#139005 Sigh... our dependency graph is a bit messed up

@kit1980
Copy link
Contributor Author
kit1980 commented Oct 27, 2024

@xuhancn Please do not edit this issue.

@kit1980
Copy link
Contributor Author
kit1980 commented Oct 27, 2024

@Skylion007 We should fix that, but 2.5.1 is an emergency patch release for specific regressions that should be completed very soon. If anyone still uses Python 3.8, they should explicitly limit the upped version of torch.

@xuhancn
Copy link
Collaborator
xuhancn commented Oct 28, 2024

@xuhancn Please do not edit this issue.

Got it, sorry.

@malfet
Copy link
Contributor
malfet commented Oct 28, 2024

See #138971 - this would be a regression between 2.5.0 and 2.5.1 and we should avoid it

@huydhn
Copy link
Contributor
huydhn commented Oct 29, 2024

All validations for 2.5.1 have been done. Thank @kit1980 @atalman @malfet

@huydhn huydhn closed this as completed Oct 29, 2024
@kiwik kiwik mentioned this issue Nov 14, 2024
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
oncall: releng In support of CI and Release Engineering
Projects
PyTorch OSS Dev Infra
Archived in project
Milestone
2.5.1
Development

No branches or pull requests
5 participants
@kit1980 @huydhn @Skylion007 @malfet @xuhancn
0