Closed krosenfeld-IDM closed 1 year ago
The specific error about Float
is super weird. Almost seems like some patch file got applied incorrectly.
I wonder if the issue is because the Julia version was not fixed in the Conda environment… could you try fixing it to 1.8? The loaded one you have is 1.9.2 which was not yet released when the experiments were ran.
In retrospect I should have fixed the entire environment rather than just the dependencies. In the meantime when I get a chance I will try to find the exact conda env.
Fixing julia to v1.8 with pysr v0.11.0 built just fine both for Docker and my local mamba environment.
Awesome. Thanks for checking! I'll make a PR that fixes this
I ran into some strange errors during the docker build stage of the
benchmark/official_competitors/pysr/Dockerfile
after building the srbench-corebenchmark/official_competitors/Dockerfile
. The error (copied below) was reproduced in mamba environments on multiple machines when installingpysr v0.11.0
using the conda-forge channel. In those environments there was no issue in installing SymbolicRegression wthin julia (e.g.,julia -e 'using Pkg; pkg"activate --temp"; pkg"add SymbolicRegression PyCall"; pkg"build PyCall"; pkg"precompile"'
)I was ultimately able to run
pysr.install()
in my mamba environments as well as build the Dockerfile by requiring pysr==0.14.3 (and runningconda clean
).I saw that there were some issues with the conda-forge installation (e.g., https://github.com/MilesCranmer/PySR/issues/347). I am not sure that this was related but am documenting for your reference.