Error when using conda environment with reticulate `incompatible architecture (have 'arm64', need 'x86_64')`

[joepowers16]

I am setting up a new M1 Mac with an arm processor evidenced by $uname -p returning "arm".

I can set up working conda environments from the terminal, but when I try to create or use a conda environment with reticulate I get the error below, which I can't make sense of. In what sense do I "need" x86_64? Isn't that an older processor?

I'll note that I can create and use virtualenv without issue in reticulate.

Error: /Users/myusername/miniconda3/envs/conda_temp2/lib/libpython3.11.dylib - dlopen(/Users/myusername/miniconda3/envs/conda_temp2/lib/libpython3.11.dylib, 0x000A): tried: '/Users/myusername/miniconda3/envs/conda_temp2/lib/libpython3.11.dylib' (mach-o file, but is an incompatible architecture (have 'arm64', need 'x86_64')), '/System/Volumes/Preboot/Cryptexes/OS/Users/jpowers4/miniconda3/envs/conda_temp2/lib/libpython3.11.dylib' (no such file), '/Users/myusername/miniconda3/envs/conda_temp2/lib/libpython3.11.dylib' (mach-o file, but is an incompatible architecture (have 'arm64', need 'x86_64'))

[michaelmayer]

I guess your R installation is using x86_64 architecture/binaries. M1 Macs will happily execute both x86_64 and arm64 binaries (though using some emulation tech for x86_64)

You can check your architecture if you run sessionInfo() within the R console.

On my M1 based Mac it reports

R version 4.3.1 (2023-06-16)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Sonoma 14.2.1
....

which clearly shows the use of arm/aarch64.

If I am upgrading to x86_64 R 4.3.2, I get from the same command

R version 4.3.2 (2023-10-31)
Platform: x86_64-apple-darwin20 (64-bit)
Running under: macOS Sonoma 14.2.1
...

which now shows x86_64.

I would suggest to try installing the arm64 version of R 4.3.2 (E.g. https://cran.r-project.org/bin/macosx/big-sur-arm64/base/R-4.3.2-arm64.pkg), make sure this version is configured in RStudio (check with sessionInfo()) and then re-try loading the conda-env again. It should work now.

Aligning the architecture of your conda env and the R installation to arm64 will definitely not only give you compatibility between the two but also the best level of performance on your M1 Mac.