How to make a python package that can have two different version of a dependency?

Question

It is now not uncommon to have a python package that is distributed in a multitude of different "flavors". This happens often with machine learning packages, e.g. onnxruntime has many "flavors" onnxruntime, onnxruntime-gpu, onnxruntime-directml, ... (same for xgboost). All those package provide the same content but have different implementations. Note that most of the time only one "flavor" can be installed in an env.

Now, let say I have a package A that depends on onnxruntime but I want to install this package (A) on a machine that has a (cuda) GPU. To take benefit of the GPU I would have to install onnxruntime-gpu instead of onnxruntime (and not both).

Ideally I would have liked to have an extra-dependency option gpu added to my package A so that if I do

pip install A

the dependency onnxruntime would be installed with it. And if I do

pip install A[gpu]

the dependency onnxruntime-gpu would be installed with it. Sadly if I do that both dependencies would be installed which is not working.

The only workable solution I see is to distribute two "flavors" of the package, A and A-gpu, which two different sets of dependencies (which I don't find ideal). Is there a better option?

Edit - Using extra markers in requirements

I also try adding condition on extra marker in my dependency list but it's not working to exclude a package to be installed while an extra is used.

With

[project.dependencies]
onnxruntime; extra!=cuda  #(or extra=="")

[project.optional-dependencies]
cuda=["onnxruntime-gpu"]

will always install onnxruntime package when installing A[cuda] because pip will always analyze the dependencies with any no extra at least one time.

That means that I can't have a valid default installation (no extra).

@kindall thanks but user guide is not addressing this specific question — MajorTom
– MajorTom, Commented Aug 13 at 14:09

J_H · Accepted Answer · 2025-08-08 15:46:36Z

1

Here is one approach that is fairly prevalent in the ecosystem. Consider these requirements.txt lines:

black[jupyter]
gymnasium[classic-control]
fastapi[standard]

jupyter is a somewhat heavyweight dependency, while black is pretty slim. Not everyone who uses it to format code will need the notebook parsing machinery; with an optional dep they don't have to pay that price.

I can imagine you might choose to add an A[gpu] dependency to one of your requirements configs.

It might even be the case that A won't install on its own -- users will have to select at least one of the optional deps, such as A[foo] or A[baz].

answered Aug 8 at 15:46

J_H

21.2k5 gold badges29 silver badges50 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Ben Reiniger Aug 11 at 23:42

The last paragraph seems closest to what's being asked for. Assuming that's correct (maybe some clarifying question-comments could help), could you elaborate how A would enforce that?

MajorTom Aug 13 at 14:26

@BenReiniger I came done to the same solution as what J_H is saying in is last paragraph (see my edit).

Collectives™ on Stack Overflow

How to make a python package that can have two different version of a dependency?

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related