FIx PR #17 and MNT: rerender

mikemhenry commented 2 years ago

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 code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
[ ] Ensured the license file is being packaged.

conda-forge-linter commented 2 years ago

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-linter commented 2 years ago

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:

noarch: python recipes are required to have a lower bound on the python version. Typically this means putting python >=3.6 in both host and run but you should check upstream for the package's Python compatibility.

conda-forge-linter commented 2 years ago

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.

mikemhenry commented 2 years ago

@hmaarrfk :wave: working on fixing some issues with pytorch packages that don't seem to pin the pytorch version, which then causes some problems: see https://github.com/pyg-team/pytorch_geometric/issues/3660

So I'm working to add some pytorch version pins to this package:

https://github.com/conda-forge/pytorch_geometric-feedstock/pull/20/files#diff-ff61408cdc05bc9667deeadb55e4aaceb1371972076b6bf6934f9008920f2bd2R5

https://github.com/conda-forge/pytorch_geometric-feedstock/pull/20/files#diff-f3725a55bf339595bf865fec73bda8ac99f283b0810c205442021f29c06eea9aR37

But I'm getting an error OSError: /home/conda/feedstock_root/build_artifacts/pytorch_geometric_1642806414629/_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehol/lib/python3.9/site-packages/torch_sparse/_convert_cpu.so: undefined symbol: _ZNK2at6Tensor6deviceEv

And it looks like pytorch_sparse also doesn't have pytorch versions pinned/a matrix setup to support multiple versions:

https://github.com/conda-forge/pytorch_sparse-feedstock/blob/master/recipe/meta.yaml

So do you know if I will need to make the same fix to that package as well?

Thanks! @jaimergp you might also know the answer to this question

hmaarrfk commented 2 years ago

This is a noarch package, which package is broken? We recently introduced a migrator and the recent PR for 1.10.1 should help alleviate future pinning problems.

hmaarrfk commented 2 years ago

Torchsparse seems broken.

hmaarrfk commented 2 years ago

it should be rebuilt.

mikemhenry commented 2 years ago

Sorry yes torchsparse was the one that was then giving me some ABI/symbols, looking at the package it didn't pin pytorch hard enough

hmaarrfk commented 2 years ago

@conda-forge-admin please restart cis

hmaarrfk commented 2 years ago

hmmm. i'm really puzzled...

hmaarrfk commented 2 years ago

@conda-forge-admin please rerender

mikemhenry commented 2 years ago

@conda-forge-admin, please rerender

mikemhenry commented 2 years ago

@conda-forge-admin, please rerender

mikemhenry commented 2 years ago

@hmaarrfk trying to get this working again our current error is https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=505349&view=logs&j=656edd35-690f-5c53-9ba3-09c10d0bea97&t=e5c8ab1d-8ff9-5cae-b332-e15ae582ed2d&l=28

'cuda_compiler_version' is undefined Which is surprising?

hmaarrfk commented 2 years ago

Are compilers needed? Why is this not a noarch?

mikemhenry commented 2 years ago

@hmaarrfk it is https://github.com/conda-forge/pytorch_geometric-feedstock/pull/20/files#diff-f3725a55bf339595bf865fec73bda8ac99f283b0810c205442021f29c06eea9aR17

But that error is coming from https://github.com/conda-forge/pytorch_geometric-feedstock/pull/20/files#diff-f3725a55bf339595bf865fec73bda8ac99f283b0810c205442021f29c06eea9aR4 so we can set pytorch {{ pytorch_version }} {{ torch_proc_type }}* correctly

mikemhenry commented 2 years ago

So I geuss the quest is, how can we make sure that when a user installs this package, they get the correct pytorch package? Since we don't have nvcc I'm guessing cuda_compiler_version doesn't exist.

hmaarrfk commented 2 years ago

The solvers should already be taking care of this.

You can simply remove the pinning.

mikemhenry commented 2 years ago

@hmaarrfk ready for review/merge

hmaarrfk commented 2 years ago

Looks good. Did you want to maintain

mikemhenry commented 2 years ago

Sure! Thanks @hmaarrfk

hmaarrfk commented 2 years ago

@conda-forge-admin please rerender

conda-forge / pytorch_geometric-feedstock

FIx PR #17 and MNT: rerender #20