How to debug pip installation error: No matching distribution found for onnxruntime

[snnn]

If when you tried to install onnxruntime python package pip said: "No matching distribution found for onnxruntime", here is how to debug it. Please follow the instructions to get more information before open an issue in Github. The title of your Github issue needs be more accurate instead of just saying "No matching distribution found for onnxruntime".

First, try to rerun the pip install command with a fixed onnxruntime version and with flags "-vv". Like this:

python3 -m pip install -vv onnxruntime==1.16.3

Or if you use a custom feed, it would be like this:

python3 -m pip install -vv --index-url=https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-12-nightly/pypi/simple/ onnxruntime==1.16.3

The command above will output a lot of things. You can redirect the log to a file and search "Skipping link: none of the wheel's tags". It will tell you what tags we provide.
Then you run " pip debug --verbose". It will tell you what tags your python supports. The intersection of the two sets shouldn't be empty. You got the error because it was empty, which means nothing we provide can satisfy your needs. So, please try to pick one tag from your python's support list that is most close to what we provide, and tell us to provide the tag. For example, you need a "win_amd64" tag but all we provide are for manylinux? Or you need a "cp312-cp312-manylinux_2_28_x86_64" tag which is for python 3.12 but we only have tags for python 3.8/3.9/3.10/3.11 ? Things like this. If you can't figure it out, give us the two lists then we will tell you. The two lists are the key information. Without them we cannot tell you why the error happened.

Besides, if your pip command has "--index-url" that points our Azure DevOPS feed, you may need to install all the dependencies of the onnxruntime package first. You may need to get the dependencies from the official pypi.org. For example, to install a nightly package from our custom Azure DevOps feed, you may need to two commands:

python -m pip install coloredlogs flatbuffers numpy packaging protobuf sympy
python -m pip install -i https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ ort_nightly

Thanks.