search

IBM / AutoMLPipeline.jl

A package that makes it trivial to create and evaluate machine learning pipeline architectures.
MIT License
355 stars 29 forks source link

Error on fresh install today #87

Closed vshesh closed 3 years ago

vshesh commented 3 years ago 
(@v.15) pkg> add AutoMLPipeline
   Building AutoMLPipeline → `~/.julia/packages/AutoMLPipeline/08lyk/deps/build.log`
┌ Error: Error building `AutoMLPipeline`:
│ FATAL ERROR: Symbol "ccalllib_libpython3.8.dylib40"not found
│ signal (6): Abort trap: 6
│ in expression starting at /Users/Vishesh/.julia/packages/AutoMLPipeline/08lyk/deps/build.jl:1
│ __pthread_kill at /usr/lib/system/libsystem_kernel.dylib (unknown line)
│ Allocations: 2543 (Pool: 2534; Big: 9); GC: 0
└ @ Pkg.Operations /Users/julia/buildbot/worker/package_macos64/build/usr/share/julia/stdlib/v1.5/Pkg/src/Operations.jl:949

System: Mac OS X: 10.13 Python version3.8 on path, through homebrew

why does automlpipeline require python? is there a way to install it without python?

ppalmes commented 3 years ago

yes, it uses sklearn models and preprocessing libraries. if you want pure julia implementation, you can use: https://github.com/IBM/AMLPipelineBase.jl

the base only supports randomforest, adaboost, and prunetree from DecisionTree.jl but you can extend it.

ppalmes commented 3 years ago

for fresh install, export this in your shell and rebuild PyCall. this is an issue in PyCall package:

from shell

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.julia/conda/3/lib export PYTHON=Conda

julia> using Pkg julia> Pkg.build("Conda") # assuming Conda is already installed julia> Pkg.build("PyCall") # assuming PyCall is already installed julia> Pkg.add("AutoMLPipeline")

ppalmes commented 3 years ago

I have mentioned this issue in the Readme