Can't install pytorch_geometric=2.3.1 & python=3.11 without a `__cuda` virtual pacakge

[mikemhenry]

$ micromamba create -n foobar --dry-run pytorch_geometric=2.3.1 python=3.11
error    libmamba Could not solve for environment specs
    The following packages are incompatible
    ├─ python 3.11**  is installable and it requires
    │  └─ python_abi 3.11.* *_cp311, which can be installed;
    └─ pytorch_geometric 2.3.1**  is uninstallable because it requires
       └─ pyg-lib 0.2.0  but there are no viable options
          ├─ pyg-lib 0.2.0 would require
          │  └─ triton   but there are no viable options
          │     ├─ triton [1.1.2|2.0.0] would require
          │     │  └─ python_abi 3.10.* *_cp310, which conflicts with any installable versions previously reported;
          │     ├─ triton 1.1.2 would require
          │     │  └─ python_abi 3.7.* *_cp37m, which conflicts with any installable versions previously reported;
          │     ├─ triton [1.1.2|2.0.0] would require
          │     │  └─ python_abi 3.8.* *_cp38, which conflicts with any installable versions previously reported;
          │     ├─ triton [1.1.2|2.0.0] would require
          │     │  └─ python_abi 3.9.* *_cp39, which conflicts with any installable versions previously reported;
          │     └─ triton 2.0.0 would require
          │        └─ pytorch * cuda* but there are no viable options
          │           ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require
          │           │  └─ python_abi 3.7.* *_cp37m, which conflicts with any installable versions previously reported;
          │           ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require
          │           │  └─ python_abi 3.8.* *_cp38, which conflicts with any installable versions previously reported;
          │           ├─ pytorch [1.10.0|1.10.1|...|1.9.1] would require
          │           │  └─ python_abi 3.9.* *_cp39, which conflicts with any installable versions previously reported;
          │           ├─ pytorch 1.11.0 would require
          │           │  └─ python_abi 3.10.* *_cp310, which conflicts with any installable versions previously reported;
          │           ├─ pytorch [1.11.0|1.12.0|...|2.0.0] would require
          │           │  └─ __cuda  , which is missing on the system;
          │           └─ pytorch [1.6.0|1.7.1|1.8.0|1.9.0|1.9.1] would require
          │              └─ python_abi 3.6.* *_cp36m, which conflicts with any installable versions previously reported;
          ├─ pyg-lib 0.2.0 would require
          │  └─ python_abi 3.10.* *_cp310, which conflicts with any installable versions previously reported;
          ├─ pyg-lib 0.2.0 would require
          │  └─ python_abi 3.8.* *_cp38, which conflicts with any installable versions previously reported;
          └─ pyg-lib 0.2.0 would require
             └─ python_abi 3.9.* *_cp39, which conflicts with any installable versions previously reported.
critical libmamba Could not solve for environment specs

I am looking into this, but wanted to make an issue to track this. Pretty sure it is an issue with pyg-lib

$ micromamba create -n foobar --dry-run pyg-lib python=3.11
error    libmamba Could not solve for environment specs
    The following packages are incompatible
    ├─ pyg-lib   is installable with the potential options
    │  ├─ pyg-lib [0.1.0|0.2.0] would require
    │  │  └─ python_abi 3.8.* *_cp38, which can be installed;
    │  ├─ pyg-lib [0.1.0|0.2.0] would require
    │  │  └─ python_abi 3.10.* *_cp310, which can be installed;
    │  ├─ pyg-lib [0.1.0|0.2.0] would require
    │  │  └─ python_abi 3.9.* *_cp39, which can be installed;
    │  ├─ pyg-lib 0.1.0 would require
    │  │  └─ pytorch [* cuda*|>=1.12.1,<1.13.0a0 ] with the potential options
    │  │     ├─ pytorch 1.12.1, which can be installed;
    │  │     ├─ pytorch [1.12.1|1.13.0|1.13.1|2.0.0] would require
    │  │     │  └─ __cuda  , which is missing on the system;
    │  │     └─ pytorch * conflicts with any installable versions previously reported;
    │  ├─ pyg-lib 0.1.0 would require
    │  │  └─ pytorch [* cuda*|>=1.13.0,<1.14.0a0 ] with the potential options
    │  │     ├─ pytorch [1.12.1|1.13.0|1.13.1|2.0.0], which cannot be installed (as previously explained);
    │  │     ├─ pytorch 1.13.0, which can be installed;
    │  │     ├─ pytorch 1.13.1, which can be installed;
    │  │     └─ pytorch * conflicts with any installable versions previously reported;
    │  ├─ pyg-lib 0.2.0 would require
    │  │  └─ pytorch [* cuda*|>=1.13.1,<1.14.0a0 ] with the potential options
    │  │     ├─ pytorch [1.12.1|1.13.0|1.13.1|2.0.0], which cannot be installed (as previously explained);
    │  │     ├─ pytorch 1.13.1, which can be installed;
    │  │     └─ pytorch * conflicts with any installable versions previously reported;
    │  └─ pyg-lib 0.2.0 would require
    │     └─ pytorch [* cuda*|>=2.0.0,<2.1.0a0 ] with the potential options
    │        ├─ pytorch [1.12.1|1.13.0|1.13.1|2.0.0], which cannot be installed (as previously explained);
    │        ├─ pytorch 2.0.0, which can be installed;
    │        └─ pytorch * conflicts with any installable versions previously reported;
    └─ python 3.11**  is uninstallable because it requires
       └─ python_abi 3.11.* *_cp311, which conflicts with any installable versions previously reported.

This output __cuda , which is missing on the system; gave me a hint that we can't install without a cuda virtual package. This was lucky since I just updated my drivers, micromamba doesn't list __cuda on my box, but with an override:

$ CONDA_OVERRIDE_CUDA="11.2" micromamba create -n foobar --dry-run pytorch_geometric=2.3.1 python=3.11 I can install it.

So is it intentional/desired behavior?

[mikemhenry]

I can install pytorch_geometric $ micromamba create -n foobar --dry-run pytorch_geometric=2.3.1 python=3.10 if I set python to 3.10

[hadim]

I removed pyg-lib as a dep so it should works now.

[mikemhenry]

Awesome! This works as expected now: CONDA_OVERRIDE_CUDA="" micromamba create -n foobar --dry-run pytorch_geometric=2.4 python=3.11