Merge branch '85-overshooting' of https://github.com/pat-alt/ECCCo.jl into 85-overshooting

experiments/grid_search.jl

@@ -4,6 +4,7 @@ using DataFrames
     grid_search(
         couterfactual_data::CounterfactualData,
         test_data::CounerfactualData;
+        warm_start::Bool = true,
         dataname::String,
         tuning_params::NamedTuple,
         kwargs...,
@@ -14,6 +15,7 @@ Perform a grid search over the hyperparameters specified by `tuning_params`. Exp
 function grid_search(
     couterfactual_data::CounterfactualData,
     test_data::CounterfactualData;
+    warm_start::Bool = true,
     dataname::String,
     n_individuals::Int = N_IND,
     tuning_params::NamedTuple,
@@ -21,19 +23,37 @@ function grid_search(
 )
 
     # Output path:
-    grid_search_path = mkpath(joinpath(DEFAULT_OUTPUT_PATH, "grid_search"))
+    grid_search_path = joinpath(DEFAULT_OUTPUT_PATH, "grid_search") 
+    if !isdir(grid_search_path)
+        mkpath(grid_search_path) 
+    end
 
     # Grid setup:
     tuning_params = [Pair.(k, vals) for (k, vals) in pairs(tuning_params)]
     grid = Iterators.product(tuning_params...)
+    n_total = length(grid)
 
     # Temporary storage on disk:
     storage_path = joinpath(grid_search_path, ".tmp_results_$(replace(lowercase(dataname), " " => "_"))")
-    mkpath(storage_path)
+    if !isdir(storage_path)
+        mkpath(storage_path)
+    end
     @info "Storing temporary results in $(storage_path)."
 
+    # Warm start:
+    if warm_start
+        existing_files = readdir(storage_path)
+        n_files = Int(floor(length(existing_files) / 2))
+        if n_files > 0
+            @info "Warm start: $(n_files) existing results found."
+            grid = Iterators.drop(grid, n_files)
+        end
+        counter = n_files + 1
+    else
+        counter = 1
+    end
+
     # Search:
-    counter = 1
     for params in grid
         @info "Running experiment $(counter)/$(length(grid)) with tuning parameters: $(params)"
 
@@ -84,7 +104,7 @@ function grid_search(
         # Deserialise:
         df_params = []
         df_outcomes = []
-        for i in 1:length(grid)
+        for i in 1:n_total
             push!(df_params, Serialization.deserialize(joinpath(storage_path, "params_$(i).jls")))
             push!(df_outcomes, Serialization.deserialize(joinpath(storage_path, "outcomes_$(i).jls")))
         end

experiments/jobscripts/tuning/generators/fmnist.sh

@@ -5,7 +5,7 @@
 #SBATCH --ntasks=10
 #SBATCH --cpus-per-task=10
 #SBATCH --partition=compute
-#SBATCH --mem-per-cpu=12GB
+#SBATCH --mem-per-cpu=16GB
 #SBATCH --account=research-eemcs-insy
 #SBATCH --mail-type=END     # Set mail type to 'END' to receive a mail when the job finishes. 
 

experiments/jobscripts/tuning/generators/mnist.sh

@@ -5,7 +5,7 @@
 #SBATCH --ntasks=10
 #SBATCH --cpus-per-task=10
 #SBATCH --partition=compute
-#SBATCH --mem-per-cpu=12GB
+#SBATCH --mem-per-cpu=16GB
 #SBATCH --account=research-eemcs-insy
 #SBATCH --mail-type=END     # Set mail type to 'END' to receive a mail when the job finishes.