uh

experiments/experiment.jl

@@ -78,7 +78,7 @@ end
 
 Run the experiment specified by `exp`.
 """
-function run_experiment(exp::Experiment; save_output::Bool=true, only_models::Bool=false)
+function run_experiment(exp::Experiment; save_output::Bool=true, only_models::Bool=ONLY_MODELS)
     
     # Setup
     @info "All results will be saved to $(exp.output_path)."

experiments/mnist.jl

-function pre_process(x; noise::Float32=0.03f0)
-    ϵ = Float32.(randn(size(x)) * noise)
-    x += ϵ
-    return x
-end
-
 # Training data:
 n_obs = 10000
 counterfactual_data = load_mnist(n_obs)
-counterfactual_data.X = pre_process.(counterfactual_data.X)
+counterfactual_data.X = ECCCo.pre_process.(counterfactual_data.X)
 
 # VAE (trained on full dataset):
 using CounterfactualExplanations.Models: load_mnist_vae

experiments/run_experiments.jl

@@ -27,18 +27,18 @@ if "circles" in datanames
     include("circles.jl")
 end
 
-# MNIST
-if "mnist" in datanames
-    @info "Running MNIST experiment."
-    include("mnist.jl")
-end
-
 # GMSC
 if "gmsc" in datanames
     @info "Running GMSC experiment."
     include("gmsc.jl")
 end
 
+# MNIST
+if "mnist" in datanames
+    @info "Running MNIST experiment."
+    include("mnist.jl")
+end
+
 if USE_MPI
     MPI.Finalize()
 end

src/ECCCo.jl

@@ -3,6 +3,7 @@ module ECCCo
 using CounterfactualExplanations
 import MLJModelInterface as MMI
 
+include("utils.jl")
 include("model.jl")
 include("sampling.jl")
 include("penalties.jl")

src/utils.jl

+function pre_process(x; noise::Float32=0.03f0)
+    ϵ = Float32.(randn(size(x)) * noise)
+    x += ϵ
+    return x
+end
\ No newline at end of file