pytorch_geometric installation and using python pkg in notebooks

[CaiYitao]

I have activated a directory and installed PythonCall, CondaPkg, installed rdkit, pytorch, but can not add pytorch_geometric

the error messages like the following(even I remove the pkgs and install other packages it still like this):

(AIforLS) pkg> conda status
CondaPkg Status /Users/mindy/百度云同步盘/StudyResearch/AIforLS/CondaPkg.toml
Not Resolved (resolve first for more information)
Packages
  rdkit
  pytorch
Channels
  pytorch
  pytorch
  conda-forge
  pytorch_geometric

(AIforLS) pkg> conda add torchvision

(AIforLS) pkg> conda resolve
    CondaPkg Found dependencies: /Users/mindy/百度云同步盘/StudyResearch/AIforLS/CondaPkg.toml
    CondaPkg Found dependencies: /Users/mindy/.julia/packages/PythonCall/XgP8G/CondaPkg.toml
    CondaPkg Creating environment
             │ /Users/mindy/.julia/artifacts/bfd9e864c155203704978fba74583859175640d2/bin/micromamba
             │ -r /Users/mindy/.julia/scratchspaces/0b3b1443-0f03-428d-bdfb-f27f9c1191ea/root
             │ create
             │ -y
             │ -p /Users/mindy/百度云同步盘/StudyResearch/AIforLS/.CondaPkg/env
             │ --override-channels
             │ --no-channel-priority
             │ python >=3.5,<4
             │ pytorch
             │ rdkit
             │ torchvision
             │ -c conda-forge
             │ -c pytorch
             └ -c pytorch_geometric

                                           __
          __  ______ ___  ____ _____ ___  / /_  ____ _
         / / / / __ `__ \/ __ `/ __ `__ \/ __ \/ __ `/
        / /_/ / / / / / / /_/ / / / / / / /_/ / /_/ /
       / .___/_/ /_/ /_/\__,_/_/ /_/ /_/_.___/\__,_/
      /_/

pytorch/noarch                                                No change
pytorch_geometric/osx-64                            12.6kB @  17.9kB/s 404 failed  0.7s
[+] 0.8s
conda-forge/osx-64       ━━━━━━━╸━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   1.3MB /  ??.?MB @   1.6MB/s  0.8s
pytorch_geometric/noarch                            12.6kB @  15.0kB/s 404 failed  0.2s
ERROR: failed process: Process(`/Users/mindy/.julia/artifacts/bfd9e864c155203704978fba74583859175640d2/bin/micromamba -r /Users/mindy/.julia/scratchspaces/0b3b1443-0f03-428d-bdfb-f27f9c1191ea/root create -y -p /Users/mindy/百度云同步盘/StudyResearch/AIforLS/.CondaPkg/env --override-channels --no-channel-priority 'python >=3.5,<4' pytorch rdkit torchvision -c conda-forge -c pytorch -c pytorch_geometric`, ProcessExited(1)) [1]

Stacktrace:
  [1] pipeline_error
    @ ./process.jl:551 [inlined]
  [2] run(::Cmd; wait::Bool)
    @ Base ./process.jl:466
  [3] run(::Cmd)
    @ Base ./process.jl:464
  [4] _run(io::IO, cmd::Cmd, args::Any; flags::Any)
    @ CondaPkg ~/.julia/packages/CondaPkg/GoPlj/src/resolve.jl:204
  [5] _resolve_conda_create(io::Any, conda_env::Any, specs::Any, channels::Any)
    @ CondaPkg ~/.julia/packages/CondaPkg/GoPlj/src/resolve.jl:145
  [6] resolve(; force::Bool, io::IO, interactive::Bool, dry_run::Bool)
    @ CondaPkg ~/.julia/packages/CondaPkg/GoPlj/src/resolve.jl:261
  [7] resolve(; force::Bool)
    @ CondaPkg.PkgREPL ~/.julia/packages/CondaPkg/GoPlj/src/PkgREPL.jl:85
  [8] resolve()
    @ CondaPkg.PkgREPL ~/.julia/packages/CondaPkg/GoPlj/src/PkgREPL.jl:85
  [9] do_cmd!(command::Pkg.REPLMode.Command, repl::REPL.LineEditREPL)
    @ Pkg.REPLMode /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/Pkg/src/REPLMode/REPLMode.jl:406
 [10] do_cmd(repl::REPL.LineEditREPL, input::String; do_rethrow::Bool)
    @ Pkg.REPLMode /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/Pkg/src/REPLMode/REPLMode.jl:387
 [11] do_cmd
    @ /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/Pkg/src/REPLMode/REPLMode.jl:378 [inlined]
 [12] (::Pkg.REPLMode.var"#24#27"{REPL.LineEditREPL, REPL.LineEdit.Prompt})(s::REPL.LineEdit.MIState, buf::IOBuffer, ok::Bool)
    @ Pkg.REPLMode /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/Pkg/src/REPLMode/REPLMode.jl:551
 [13] #invokelatest#2
    @ ./essentials.jl:729 [inlined]
 [14] invokelatest
    @ ./essentials.jl:727 [inlined]
 [15] run_interface(terminal::REPL.Terminals.TextTerminal, m::REPL.LineEdit.ModalInterface, s::REPL.LineEdit.MIState)
    @ REPL.LineEdit /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/REPL/src/LineEdit.jl:2509
 [16] run_frontend(repl::REPL.LineEditREPL, backend::REPL.REPLBackendRef)
    @ REPL /Applications/Julia-1.8.app/Contents/Resources/julia/share/julia/stdlib/v1.8/REPL/src/REPL.jl:1248
 [17] (::REPL.var"#49#54"{REPL.LineEditREPL, REPL.REPLBackendRef})()
    @ REPL ./task.jl:476

2. even though installed rdkit and pytorch in this directory, still can not use it in Pluto(I know it has its own temp directory, I install the rdkit in Pluto successfully.

) or Jupyter notebook, but can be called in julia file, what did I missing?

I use VSCODE for Julia and jupyter.

Thank you!

[cjdoris]

You have pytorch_geometric as a channel, but it should be a package.

[CaiYitao]

Yeah, I install the pytorch_geometric, when resolving it failed, I removed it, but it was added to the channel automatically, how can I remove it in Channel, I tried conda remove/rm/rm_channel/channel_rm not working.

BTW, What could be the problem I can not import rdkit in Pluto, even it should be installed successfully.

[icweaver]

Can CondaPkg add multiple pkgs at the same time?

I'm also interested about this. I see that we can do pkg> conda add a b c in the REPL, but am not sure what the equivalent would be with CondaPkg. I guess we could broadcast it:

CondaPkg.add.(("a", "b", "c"))

but idk if doing multiple calls to add is the same as passing a vector of packages in conda/(micro)mamba?

[cjdoris]

Passing a vector should work, like

CondaPkg.add(["a", "b", "c"])

@icweaver's solution also works, but CondaPkg now resolves things immediately so you'll end up doing conda install three times. To avoid that do

CondaPkg.add.(["a", "b", "c"], resolve=false)
CondaPkg.resolve()

[icweaver]

Passing a vector should work

Great! I was hoping that there was a way like this already. It currently seems to throw an error for me on CondaPkg v0.2.10 and main (78f2a63) though:

julia> using CondaPkg

julia> CondaPkg.add(["matplotlib", "numpy"])
ERROR: MethodError: no method matching add!(::Dict{String, Any}, ::String)
Closest candidates are:
  add!(::Any, ::CondaPkg.PkgSpec) at ~/.julia/packages/CondaPkg/Olo5Y/src/deps.jl:197
  add!(::Any, ::CondaPkg.ChannelSpec) at ~/.julia/packages/CondaPkg/Olo5Y/src/deps.jl:214
  add!(::Any, ::CondaPkg.PipPkgSpec) at ~/.julia/packages/CondaPkg/Olo5Y/src/deps.jl:219
Stacktrace:
 [1] add(pkgs::AbstractVector; resolve::Any)
   @ CondaPkg ~/.julia/packages/CondaPkg/Olo5Y/src/deps.jl:187
 [2] add(pkgs::AbstractVector)
   @ CondaPkg ~/.julia/packages/CondaPkg/Olo5Y/src/deps.jl:185
 [3] top-level scope
   @ REPL[5]:1

[CaiYitao]

yeah, me too, got the same error

[cjdoris]

Ah yeah adding or rming a vector of things requires the things to be PkgSpecs, ChannelSpecs or PipPkgSpecs, e.g.

CondaPkg.add([CondaPkg.PkgSpec("matplotlib"), CondaPkg.PkgSpec("numpy")])

But this is not API (yet) and TBH I'd just do this instead:

CondaPkg.add("matplotlib", resolve=false)
CondaPkg.add("numpy", resolve=false)
CondaPkg.resolve()