Try to update to 2.1.0 and python 3.12

[hmaarrfk]

All linux logs:
log_files.zip

Checklist

• Used a personal fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

• Selectors are suggested to take a <two spaces>#<one space>[<expression>] form. See lines [26]

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[hmaarrfk]

@xhochy can i make it so that all the linux builds timeout? I can build all of those myself.

[hmaarrfk]

while the osx is anticipated to pass, the linux builds will take longer since i'm running them on a single machine.

[hmaarrfk]

log_files.zip

[hmaarrfk]

Any objections to merging this?

[Tobias-Fischer]

Not from me, this is very exciting, thanks a lot for all your efforts, @hmaarrfk!!!

[jakirkham]

Thanks Mark! 🙏

[jakirkham]

Sorry had one follow up question. Did we build packages for CUDA 12 as well? Or are those pending?

[hmaarrfk]

they half exist..... the pytorch-gpu is built, but the pytorch seems to be missing...

[hmaarrfk]

linux_64_blas_implmklc_compiler_version12cuda_compilercuda-nvcccuda_compiler_version12.0cxx_compiler_version12numpy1.26python3.12.____cpython-log.txt

[hmaarrfk]

sorry, they are broken. i should have looked more closely. it is hard with the sheer numberof builds....

[hmaarrfk]

  -- Unable to find cublas_v2.h in either "/home/conda/feedstock_root/build_artifacts/pytorch-recipe_1699186934304/_build_env/targets/x86_64-linux/include" or "/home/conda/feedstock_root/build_artifacts/pytorch-recipe_1699186934304/_build_env/math_libs/include"
  -- Found CUDAToolkit: /home/conda/feedstock_root/build_artifacts/pytorch-recipe_1699186934304/_build_env/targets/x86_64-linux/include (found version "12.0.76")
  -- Performing Test CMAKE_HAVE_LIBC_PTHREAD
  -- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
  -- Check if compiler accepts -pthread
  -- Check if compiler accepts -pthread - yes
  -- Found Threads: TRUE
  CMake Error at cmake/public/cuda.cmake:65 (message):
    Found two conflicting CUDA installs:

    V12.0.76 in
    '/home/conda/feedstock_root/build_artifacts/pytorch-recipe_1699186934304/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placeh/targets/x86_64-linux/include'
    and

    V12.0.76 in
    '/home/conda/feedstock_root/build_artifacts/pytorch-recipe_1699186934304/_build_env/targets/x86_64-linux/include'
  Call Stack (most recent call first):
    cmake/Dependencies.cmake:44 (include)
    CMakeLists.txt:722 (include)

[jakirkham]

Completely understand

Thanks Mark! 🙏

Interesting IIUC this worked in 2.0.0 and then started failing with 2.1.0. Wonder what caused the change

[jakirkham]

Think this line might be part of the issue

pytorch-cpu-feedstock/recipe/build_pytorch.sh

Line 155 in dc47045

export CUDA_TOOLKIT_ROOT_DIR=$CUDA_HOME

$CUDA_HOME is no longer defined in CUDA 12. Think if we don't set this in the CUDA 12 case, the logic already baked into cuda-nvcc for CMake should take affect

If that doesn't work for some reason, am happy to discuss alternatives

[jakirkham]

cc @RaulPPelaez (for awareness)

[hmaarrfk]

Interesting IIUC this worked in 2.0.0 and then started failing with 2.1.0. Wonder what caused the change

I feel like it also worked when I tested things for 2.1.0, but then eventually started to stop working...

[jakirkham]

More discussion in PR: #203