Failing test for bottom-up vtree learning

[guyvdbroeck]

Under Julia 1.6rc2, there are tests under https://github.com/Juice-jl/ProbabilisticCircuits.jl/blob/master/test/structurelearner/vtree_learner_tests.jl that fail, because the learned bottom-up vtree is different from the one hardcoded in the test:

julia> print_tree(v)
━PlainVtreeInnerNode(1,2,3,4,5,6,7,8,9,10)
 ┣━PlainVtreeInnerNode(2,4,6,8,9,10)
 ┃ ┣━PlainVtreeInnerNode(2,8,9,10)
 ┃ ┃ ┣━PlainVtreeInnerNode(2,8)
 ┃ ┃ ┃ ┣━PlainVtreeLeafNode(2)
 ┃ ┃ ┃ ┗━PlainVtreeLeafNode(8)
 ┃ ┃ ┗━PlainVtreeInnerNode(9,10)
 ┃ ┃   ┣━PlainVtreeLeafNode(9)
 ┃ ┃   ┗━PlainVtreeLeafNode(10)
 ┃ ┗━PlainVtreeInnerNode(4,6)
 ┃   ┣━PlainVtreeLeafNode(4)
 ┃   ┗━PlainVtreeLeafNode(6)
 ┗━PlainVtreeInnerNode(1,3,5,7)
   ┣━PlainVtreeInnerNode(1,7)
   ┃ ┣━PlainVtreeLeafNode(1)
   ┃ ┗━PlainVtreeLeafNode(7)
   ┗━PlainVtreeInnerNode(3,5)
     ┣━PlainVtreeLeafNode(3)
     ┗━PlainVtreeLeafNode(5)
julia> p1
2-element Vector{Vector{Vector{T} where T}}:
 [[[1, 7], [9, 10]], [4, 6]]
 [[2, 8], [3, 5]]

It's unclear whether this is just randomness of ties (perhaps a change in one of the dependencies) or whether there is a bug.

[guyvdbroeck]

closed by 64c3004
bottom-up vtree learning now requires the user to manually import BlossomV