site_energy_d gives back MethodError! about julip.jl

[cortner]

can you please produce a complete script to reproduce it + post you Project configuration via ] status

[cortner]

It's not unlikely that this just hasn't been implemented because it was never used. What is your use-case? Fitting to site-entropy derivatives?

[tinatorabi]

Yes, I need it for fitting to site entropy derivatives. here is the script:

using JuLIP, ASE, ACEpotentials
cu_bulk= bulk(:Cu, cubic=true, pbc=true) * 2
eam = JuLIP.Potentials.EAM("Cu.eam.alloy")

r0 = 2.55
basis = ACE1x.ace_basis(elements = [:Cu],
                  order = 3,
                  totaldegree = 10, 
                  rcut = 5.506786, 
                  r0 = r0,)
se= site_energy_d(basis, cu_bulk,1)

EAM potential file: https://www.ctcms.nist.gov/potentials/Download/2001--Mishin-Y-Mehl-M-J-Papaconstantopoulos-D-A-et-al--Cu-1/2/Cu01.eam.alloy

(@v1.9) pkg> status
Status `~/.julia/environments/v1.9/Project.toml`
  [8c4e8d19] ACE1pack v0.5.0
  [3b96b61c] ACEpotentials v0.6.3
  [51974c44] ASE v0.5.4
  [7d9fca2a] Arpack v0.5.4
  [6e4b80f9] BenchmarkTools v1.3.2
  [a93c6f00] DataFrames v1.6.1
  [31c24e10] Distributions v0.25.100
  [b305315f] Elliptic v1.0.1
  [6a86dc24] FiniteDiff v2.21.1
  [f6369f11] ForwardDiff v0.10.36
  [38e38edf] GLM v1.9.0
  [7073ff75] IJulia v1.24.2
  [945c410c] JuLIP v0.14.4
  [b964fa9f] LaTeXStrings v1.3.0
  [10e44e05] MATLAB v0.8.4
  [6f286f6a] MultivariateStats v0.10.2
  [2fcf5ba9] NeighbourLists v0.5.5
  [429524aa] Optim v1.7.7
  [91a5bcdd] Plots v1.39.0
  [438e738f] PyCall v1.96.1
  [d330b81b] PyPlot v2.11.2
  [37e2e3b7] ReverseDiff v1.15.1
  [47a9eef4] SparseDiffTools v2.6.0
  [684fba80] SparsityDetection v0.3.4
  [90137ffa] StaticArrays v1.6.3
  [0c5d862f] Symbolics v5.5.3

[cortner]

It seems there were multiple potential bugs but not too much trouble to fix them.

@tinetorabi - can you please update your environment and make sure the you have [email protected] and [email protected] and then try again? If it works, then you can close this issue.

One comment : using site_energy and site_energy_d is actually an extremely inefficient way of fitting to per/atom quantities. Your code will spend about 99% of its time in allocating memory and then garbage-collecting it. No harm to try, but don't be surprised if it's slow. I will then show you in our next meeting how to do this "correctly".

[cortner]

It seems there were multiple potential bugs but not too much trouble to fix them.

@tinetorabi - can you please update your environment and make sure the you have [email protected] and [email protected] and then try again? If it works, then you can close this issue.

One comment : using site_energy and site_energy_d is actually an extremely inefficient way of fitting to per/atom quantities. Your code will spend about 99% of its time in allocating memory and then garbage-collecting it. No harm to try, but don't be surprised if it's slow. I will then show you in our next meeting how to do this "correctly".