cavalab/srbench

deprecated and broken algorithms

Opened this issue · 5 comments

lacava commented

Since PR #151 (commit 3b7181f), the following algorithms are either considered broken or deprecated:

  • gsgp (no plan to fix)
  • mrgp (no plan to fix)
  • sbp-gp (no plan to fix)
  • nsga-dcp (no plan to fix)
  • taylorgp (no plan to fix)
  • ai-feynman
  • dsr
  • uDSR
  • dagsearch (cc @kahlmeyer94)

the code for these algs should be moved to a separate branch and the authors invited to update them so that they work with the current version of the benchmark.

kahlmeyer94 commented

For dagsearch, it looked like an issue with scikit-learn. I think I fixed it now (at least it is working on my machine now).
Could you rerun the build check for dagsearch please?

Thanks :)

kahlmeyer94 commented

@lacava Im really confused as for where the error is. Could you help me find and fix it?
Is it because of this error message
AttributeError: module 'pkgutil' has no attribute 'ImpImporter'. Did you mean: 'zipimporter'?

or because it couldnt find torch?
Searching for torch Reading https://pypi.org/simple/torch/ Error: Process completed with exit code 1.

Edit: It looks like Pytorch is not yet available for python 3.12. (see here).
Whenever this is the case, the current version should work.
Alternatively one could use python 3.11.