Try to allow other build numbers of nifty and protobuf to be installed

[conda-forge-admin]

This creates 2 "noarch packages" to hack around the limitations that conda-forge does not have a pytorch package.

• The one built by the linux runner -- limited to unix platforms. loosens the nifty pin
• The one built by the OSX runner -- not limited to unix. -- limits the nifty pin to something very strict.
• I’ve also loosened the protobuf requirement. This is only a problem for users that mix conda-forge and default channel (which is not supported) and the PyTorch channel is compatible with the default channel, not conda-forge.
• I've validated that both packages have different hash numbers.

This is pretty "not ok" at conda-forge and I would recommend you don't give these kinds of hack instructions to your end users.

You can see my "comments" on the status of pytorch at conda-forge:
conda-forge/conda-forge.github.io#2381

-- hmaarrfk cc @hmaarrfk

[conda-forge-admin]

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/meta.yaml) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe/meta.yaml:

• ℹ️ noarch: python recipes should usually follow the syntax in our documentation for specifying the Python version.
  • For the test.requires section of the recipe, you should usually use python {{ python_min }} for the python entry.
  • If the package requires a newer Python version than the currently supported minimum version on conda-forge, you can override the python_min variable by adding a Jinja2 set statement at the top of your recipe (or using an equivalent context variable for v1 recipes).

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/12423580550. Examine the logs at this URL for more detail.}

[hmaarrfk]

@conda-forge-admin please rerender

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you but ran into some issues. Please check the output logs of the GitHub Actions workflow below for more details. You can also ping conda-forge/core (using the @ notation) for further assistance or you can try rerendering locally.

The following suggestions might help debug any issues:

• Is the recipe/{meta.yaml,recipe.yaml} file valid?
• If there is a recipe/conda-build-config.yaml file in the feedstock make sure that it is compatible with the current global pinnnings.
• Is the fork used for this PR on an organization or user GitHub account? Automated rerendering via the webservices admin bot only works for user GitHub accounts.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/12420255842. Examine the logs at this URL for more detail.}

[hmaarrfk]

hmm we got much stricter in even ensuring that things can be solved...

[constantinpape]

How would a user specify which of the two packages to install? Would they need to specify the exact hashes? Would there be a way to have a human readable way?
Something like micro_sam (the default for unix-like without strict nifty pin) and micro_sam-win (the stricter windows pin). Or would we need completely separate packages for this?

[constantinpape]

Just for me to understand: you are using osx here to build with a stricter pin because it would not build on win due to missing pytorch. Is that correct?

[hmaarrfk]

yes. when i tried to build with win it didn't work so well:
https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=1115800&view=logs&j=fe7faf59-69bb-5487-b8fd-a35a7bda3a51&t=55731638-dd49-5144-2129-5101b9dd1960

this solution is in fact, very hacky...

[hmaarrfk]

How would a user specify which of the two packages to install?

They don't. This really isn't the conda(-forge) way.

Would they need to specify the exact hashes?

yes.

Would there be a way to have a human readable way?

Similar to https://github.com/computational-cell-analytics/micro-sam/blob/eb0def46d54a8e26758df05b5236f58fde8b56b9/workshops/i2k_2024/README.md?plain=1#L41C10-L41C85
just add

conda install -c pytorch -c conda-forge "micro_sam>=1.1" "pytorch>=2.4" "protobuf<5" cpuonly "nifty=1.2.1=*_4"

(I can't say this enough, this is very fragile for many reasons).

But specifying this is not necessary:

1. Linux + OSX users can get a working compatible pytorch from conda-forge
2. Windows users will get limited to the nifty=1.2.1=*_4 and protobuf<5

Something like micro_sam (the default for unix-like without strict nifty pin) and micro_sam-win (the stricter windows pin). Or would we need completely separate packages for this?

Conda-forge is not the right place to make hacks for this kind of thing. You fork this feedstock, use the CIs to upload whatever package you want to your own channel, but please don't do this kind of stuff at conda-forge.

The hacks here are at the border of acceptable and may be deemed too much by other core members already...

[constantinpape]

Hi @hmaarrfk ,
thanks for the feedback here. Given this, I would opt for a cleaner solution, I summarized here:
computational-cell-analytics/micro-sam#813

Let me know what you think (in that issue) and then I would start with implementing it later.

[constantinpape]

I am closing this in favor of #26 . Thanks for the feedback on this!