Merge pull request #165 from CliMA/ne/hotfix

Fix failure_rate kwarg in WorkerBackend, improve error message

Author: nefrathenrici

Project.toml

@@ -1,7 +1,7 @@
 name = "ClimaCalibrate"
 uuid = "4347a170-ebd6-470c-89d3-5c705c0cacc2"
 authors = ["Climate Modeling Alliance"]
-version = "0.0.14"
+version = "0.0.15"
 
 [deps]
 Distributed = "8ba89e20-285c-5b6f-9357-94700520ee1b"

docs/literate_example.jl

@@ -111,7 +111,7 @@ end
 # for the ensemble update step. The observation map just takes in the iteration number,
 # and always outputs an array. 
 # For observation map output `G_ensemble`, `G_ensemble[:, m]` must the output of ensemble member `m`.
-# This is needed for compatability with EnsembleKalmanProcesses.jl.
+# This is needed for compatibility with EnsembleKalmanProcesses.jl.
 const days = 86_400
 function CAL.observation_map(iteration)
     single_member_dims = (1,)

experiments/surface_fluxes_perfect_model/Manifest.toml

@@ -202,7 +202,7 @@ weakdeps = ["SparseArrays"]
 deps = ["Distributed", "Distributions", "EnsembleKalmanProcesses", "JLD2", "Logging", "Random", "TOML", "YAML"]
 path = "../.."
 uuid = "4347a170-ebd6-470c-89d3-5c705c0cacc2"
-version = "0.0.13"
+version = "0.0.14"
 
     [deps.ClimaCalibrate.extensions]
     CESExt = "CalibrateEmulateSample"

src/backends.jl

@@ -237,7 +237,10 @@ const DEFAULT_FAILURE_RATE = 0.5
 Run a full calibration on the given backend.
 
 If the EKP struct is not given, it will be constructed upon initialization. 
-The experiment configuration (ensemble size, prior, observations, etc) can be 
+While EKP keyword arguments are passed through to the EKP constructor, if using
+many keywords it is recommended to construct the EKP object and pass it into `calibrate`.
+
+The experiment configuration (ensemble size, prior, observations, etc) can be
 wrapped in an ExperimentConfig or passed in as arguments to the function.
 
 Available Backends: WorkerBackend, CaltechHPCBackend, ClimaGPUBackend, DerechoBackend, JuliaBackend
@@ -286,14 +289,22 @@ function calibrate(
     worker_pool = default_worker_pool(),
     ekp_kwargs...,
 )
-    eki = ekp_constructor(
+    ekp = ekp_constructor(
         ensemble_size,
         prior,
         observations,
         noise;
         ekp_kwargs...,
     )
-    return calibrate(b, eki, n_iterations, prior, output_dir; worker_pool)
+    return calibrate(
+        b,
+        ekp,
+        n_iterations,
+        prior,
+        output_dir;
+        failure_rate,
+        worker_pool,
+    )
 end
 
 function calibrate(

src/workers.jl

@@ -17,8 +17,6 @@ function run_worker_iteration(
     worker_pool = default_worker_pool(),
     failure_rate = DEFAULT_FAILURE_RATE,
 )
-    # Create a channel to collect results
-    results = Channel{Any}(ensemble_size)
     nfailures = 0
     @sync begin
         for m in 1:(ensemble_size)
@@ -39,8 +37,10 @@ function run_worker_iteration(
     end
     iter_failure_rate = nfailures / ensemble_size
     if iter_failure_rate > failure_rate
-        error(
-            "Ensemble for iter $iter had a $(iter_failure_rate * 100)% failure rate",
+        throw(
+            ErrorException(
+                "Execution halted: Iteration $iter had a $(round(iter_failure_rate * 100; digits=2))% failure rate, exceeding the maximum allowed threshold of $(failure_rate * 100)%.",
+            ),
         )
     end
 end