Read on science channel about :

« The offending upstream commit is here: github.com/tensorflow/tensorflow@e85860e838

There is no way to build Tensorflow without a bundled Python toolchain. The only way I see is to trick it into "downloading" our provided Python toolchain from a file:// URL.

I find this battle against upstream's approach towards reproducibility exhausting. »

🤔 Could you stop to say: “yeah Guix is great but I don’t use because it doesn’t provide what I need”…

…and instead explicitly say: “my research outcome is more important than doing it correctly”.

https://codeberg.org/guix-science/guix-science/issues/593#issuecomment-11048234