DiffEqFlux does not expose sciml_train

Specific Domains Machine Learning
1

I’m trying to get DiffEqFlux.jl 1.4.5 running on Julia 1.6.5 (Mac x86). When running the Lotka-Volterra example in the docs verbatim, it throws the following error:

UndefVarError: sciml_train not defined

Stacktrace:
 [1] getproperty(x::Module, f::Symbol)
   @ Base ./Base.jl:26
 [2] top-level scope
   @ In[191]:21
 [3] eval
   @ ./boot.jl:360 [inlined]
 [4] include_string(mapexpr::typeof(REPL.softscope), mod::Module, code::String, filename::String)
   @ Base ./loading.jl:1116

So far, I have tried

  • downgrading DiffEqFlux down to 1.4.0
  • installing in a temporary environment
  • running in a Julia 1.7 setup

This is what’s installed in the current environment:

julia> Pkg.status()
      Status `~/.julia/environments/v1.6/Project.toml`
  [c52e3926] Atom v0.12.36
  [336ed68f] CSV v0.10.2
  [5ae59095] Colors v0.12.8
  [a93c6f00] DataFrames v1.3.2
  [7806a523] DecisionTree v0.10.11
  [aae7a2af] DiffEqFlux v1.45.0
  [1130ab10] DiffEqParamEstim v1.23.0
  [41bf760c] DiffEqSensitivity v6.69.1
  [0c46a032] DifferentialEquations v7.1.0
  [587475ba] Flux v0.12.9
  [c91e804a] Gadfly v1.3.4
  [a75be94c] GalacticOptim v2.3.0
  [f67ccb44] HDF5 v0.16.2
  [7073ff75] IJulia v1.23.2
  [a09fc81d] ImageCore v0.9.3
  [e5e0dc1b] Juno v0.8.4
  [50d2b5c4] Lazy v0.15.1
  [eb30cadb] MLDatasets v0.5.15
  [0db19996] NBInclude v2.3.0
  [429524aa] Optim v1.6.1
  [1dea7af3] OrdinaryDiffEq v6.6.6
  [d96e819e] Parameters v0.12.3
  [91a5bcdd] Plots v1.25.8
  [438e738f] PyCall v1.93.0
  [ce6b1742] RDatasets v0.7.7
  [295af30f] Revise v3.3.1
  [3646fa90] ScikitLearn v0.6.4
  [2913bbd2] StatsBase v0.33.14
  [9f7883ad] Tracker v0.2.18
  [e88e6eb3] Zygote v0.6.34

Any thoughts?

2

If you’re actually on v1.45 it should be fine. Did you start a new Julia session after updating the current environment? You might’ve had old stuff loaded.

3

Alright: cold reboot worked. Strange, though - I opened and closed several REPLs, Atom and Jupyter instances and that didn’t change a thing.

Anyway, works just fine now, appreciate the quick response!