log_ml_estimate is NaN

[marleneberke]

julia> include("April_May/execute_Mario_model.jl")

percept 1
log_ml_estimate is 0.0
ess at start of loop is 100.00000000000004
ess after perturbation is 100.00000000000004
effective sample size: 100.00000000000004, doing resample: false
ess after resample is 100.00000000000004
ess after particle filter step is NaN
percept 2
log_ml_estimate is NaN
ess at start of loop is NaN
ERROR: LoadError: ArgumentError: Categorical: the condition isprobvec(p) is not satisfied.
Stacktrace:
[1] macro expansion at /Users/marleneberke/.julia/packages/Distributions/dTXqn/src/utils.jl:6 [inlined]
[2] #_#26 at /Users/marleneberke/.julia/packages/Distributions/dTXqn/src/univariate/discrete/categorical.jl:30 [inlined]
[3] #Categorical#27 at /Users/marleneberke/.julia/packages/Distributions/dTXqn/src/univariate/discrete/categorical.jl:34 [inlined]
[4] DiscreteNonParametric at /Users/marleneberke/.julia/packages/Distributions/dTXqn/src/univariate/discrete/categorical.jl:34 [inlined]
[5] random at /Users/marleneberke/.julia/packages/Gen/zBy5S/src/modeling_library/categorical.jl:21 [inlined]
[6] Categorical at /Users/marleneberke/.julia/packages/Gen/zBy5S/src/modeling_library/categorical.jl:25 [inlined]
[7] sample_unweighted_traces(::Gen.ParticleFilterState{Gen.DynamicDSLTrace}, ::Int64) at /Users/marleneberke/.julia/packages/Gen/zBy5S/src/inference/particle_filter.jl:67
[8] particle_filter(::Int64, ::Array{Any,1}, ::Gen.DynamicDSLChoiceMap, ::Int64) at /Users/marleneberke/Documents/03_Yale/Projects/001_Mask_RCNN/ORB_project3/April_May/inference_Mario_model.jl:194
[9] top-level scope at ./util.jl:308 [inlined]
[10] top-level scope at /Users/marleneberke/Documents/03_Yale/Projects/001_Mask_RCNN/ORB_project3/April_May/execute_Mario_model.jl:0
[11] include(::String) at ./client.jl:439
[12] top-level scope at REPL[80]:1
in expression starting at /Users/marleneberke/Documents/03_Yale/Projects/001_Mask_RCNN/ORB_project3/April_May/execute_Mario_model.jl:154

[marleneberke]

@belledon I know I've seen this before. Last time, I think it was the traces had probability 0 or something