MilesCranmer/PySR: v0.17.0

What's Changed

• Bump docker/build-push-action from 3 to 5 by @dependabot in https://github.com/MilesCranmer/PySR/pull/510
• Bump actions/cache from 3 to 4 by @dependabot in https://github.com/MilesCranmer/PySR/pull/526
• Update colab notebook to use juliaup by @MilesCranmer in https://github.com/MilesCranmer/PySR/pull/531
• Bump peter-evans/create-pull-request from 5 to 6 by @dependabot in https://github.com/MilesCranmer/PySR/pull/539
• (BREAKING) Rewrite Julia interface with PyJulia -> JuliaCall; other changes by @MilesCranmer @cjdoris @mkitti in https://github.com/MilesCranmer/PySR/pull/535

Detailed changes from #535

• (BREAKING) Changed PyJulia with JuliaCall
  • Need to change eval -> seval
  • Manually converting to Vector when calling SymbolicRegression.jl functions (otherwise would get passed as PyList{Any}; see https://github.com/JuliaPy/PythonCall.jl/issues/441)
  • Wrapped equation_search code with jl.PythonCall.GC.disable() to avoid multithreading-related segfaults (https://github.com/JuliaPy/PythonCall.jl/issues/298)
  • Manually convert np.str_ to str before passing to variable_names, otherwise it becomes a PyArray and not a String (might be worth adding a workaround, it seems like PyJulia does this automatically)
• (BREAKING) Julia is now installed automatically when you import pysr (via JuliaCall)
• (BREAKING) The user no longer needs to run python -m pysr install. The install process is done by JuliaCall at import time.
  • Removed code related to pysr.install() and python -m pysr install because JuliaCall now handles this.
  • python -m pysr install will not give a warning and do nothing.
• (BREAKING) Remove the feynman problems dataset. Didn't seem good to have a dataset within a library itself.
• (BREAKING) Deprecated julia_project argument (ignored; no effect). The user now needs to set this up by customizing juliapkg.json. See updated documentation for instructions.
• (BREAKING) Switch from python -m pysr.test [test] to python -m pysr test [test].
• Switches to pyproject.toml for building rather than setup.py. However, setup.py install should still work.
• Dependencies are now managed by pyjuliapkg rather than the custom code we made. Simplifies things a lot!
• Rather than storing the raw julia variables in PySRRegressor, I am now storing a serialized version of them. This means you can now pickle the search state and warm-start the search from a file, in another Python process!
  • Not breaking! Because self.raw_julia_state_ will deserialize it automatically for you
• SymbolicRegression is now available to import from PySR:

from pysr import SymbolicRegression as SR
x1 = SR.Node(feature=1)  # Create expressions manually

• SymbolicRegression options are accessible in <model>.julia_options_ (generated from a serialized format for pickle safety) so that the user can call a variety of functions in SymbolicRegression.jl directly.
• Deprecated various kwargs to match SymbolicRegression.jl (old names will still work, so this is not breaking):
  • ncyclesperiteration => ncycles_per_iteration
  • loss => elementwise_loss
  • full_objective => loss_function
• Fixes Jupyter printing by automatically loading the juliacall.ipython extension at import time
• Adds Zygote.jl to environment by default
• Does unittesting on an example Jupyter notebook

Full Changelog: https://github.com/MilesCranmer/PySR/compare/v0.16.9...v0.17.0