Clean up slurm pipeline (#73)

docs/src/experiment_setup_guide.md

@@ -136,7 +136,7 @@ observation_map(::Val(Symbol(experiment_id)), iteration)
 This function must load in model diagnostics for each ensemble member in the iteration and construct an array `arr = Array{Float64}(undef, dims..., ensemble_size)` such that
 `arr[:, i]` will return the i-th ensemble member's observation map output. Note this floating point precision is required for the EKI update step.
 
-In the update step of EKI, the array will be saved in a JLD2 file named `observation_map.jld2` in the iteration folder of the output directory.
+In the update step of EKI, the array will be saved in a JLD2 file named `G_ensemble.jld2` in the iteration folder of the output directory.
 
 As an example, in `observation_map(iteration)` in the `sphere_held_suarez_rhoe_equilmoist` experiment, we have the following sequence:
 

experiments/sphere_held_suarez_rhoe_equilmoist/generate_observations.sbatch

@@ -3,7 +3,7 @@
 #SBATCH --ntasks=64
 #SBATCH --cpus-per-task=8
 #SBATCH --partition=expansion
-#SBATCH --output="experiments/sphere_held_suarez_rhoe_equilmoist/truth_simulation/model_log.out"
+#SBATCH --output="experiments/sphere_held_suarez_rhoe_equilmoist/truth_simulation/model_log.txt"
 #SBATCH --partition=expansion
 
 # Configure the environment

experiments/sphere_held_suarez_rhoe_equilmoist/observation_map.jl

@@ -3,17 +3,15 @@ using Statistics
 import YAML
 import EnsembleKalmanProcesses: TOMLInterface
 import JLD2
-import CalibrateAtmos: observation_map
+import CalibrateAtmos: observation_map, get_ekp_config
 using ClimaAnalysis
 export observation_map
 
 function observation_map(::Val{:sphere_held_suarez_rhoe_equilmoist}, iteration)
     experiment_id = "sphere_held_suarez_rhoe_equilmoist"
-    config =
-        YAML.load_file(joinpath("experiments", experiment_id, "ekp_config.yml"))
+    config = get_ekp_config(experiment_id)
     output_dir = config["output_dir"]
     ensemble_size = config["ensemble_size"]
-    model_output = "ta_60d_average.nc"
 
     dims = 1
     G_ensemble = Array{Float64}(undef, dims..., ensemble_size)

experiments/surface_fluxes_perfect_model/Manifest.toml

@@ -1,6 +1,6 @@
 # This file is machine-generated - editing it directly is not advised
 
-julia_version = "1.10.1"
+julia_version = "1.10.2"
 manifest_format = "2.0"
 project_hash = "8b52a1f87337958a3f0ec6e731cccb862b78de20"
 
@@ -300,7 +300,7 @@ uuid = "83423d85-b0ee-5818-9007-b63ccbeb887a"
 version = "1.16.1+1"
 
 [[deps.CalibrateAtmos]]
-deps = ["CalibrateEmulateSample", "ClimaComms", "ClimaCore", "ClimaParams", "Distributions", "EnsembleKalmanProcesses", "JLD2", "PrecompileTools", "Random", "SciMLBase", "TOML", "YAML"]
+deps = ["CalibrateEmulateSample", "ClimaComms", "ClimaParams", "Distributions", "EnsembleKalmanProcesses", "JLD2", "PrecompileTools", "Random", "SciMLBase", "TOML", "YAML"]
 path = "../.."
 uuid = "4347a170-ebd6-470c-89d3-5c705c0cacc2"
 version = "0.1.0"

experiments/surface_fluxes_perfect_model/generate_truth.jl

@@ -1,7 +1,5 @@
 # generate_truth: generate true y, noise and x_inputs
-using Pkg
 experiment_id = "surface_fluxes_perfect_model"
-Pkg.activate("experiments/$experiment_id")
 
 import SurfaceFluxes as SF
 import SurfaceFluxes.Parameters as SFPP

experiments/surface_fluxes_perfect_model/observation_map.jl

@@ -2,7 +2,7 @@ using Statistics
 import YAML
 import EnsembleKalmanProcesses: TOMLInterface
 import JLD2
-import CalibrateAtmos: observation_map
+import CalibrateAtmos: observation_map, get_ekp_config
 
 """
     observation_map(::Val{:surface_fluxes_perfect_model}, iteration)
@@ -12,8 +12,7 @@ as specified by process_member_data, for the given iteration.
 """
 function observation_map(::Val{:surface_fluxes_perfect_model}, iteration)
     experiment_id = "surface_fluxes_perfect_model"
-    config =
-        YAML.load_file(joinpath("experiments", experiment_id, "ekp_config.yml"))
+    config = get_ekp_config(experiment_id)
     output_dir = config["output_dir"]
     ensemble_size = config["ensemble_size"]
     model_output = "model_ustar_array.jld2"

pipeline.sh

@@ -13,6 +13,7 @@ fi
 init_id=$(sbatch --parsable \
                  --output=$logfile \
                  --partition=$partition \
+                 --export=generate_data=$generate_data \
                  slurm/initialize.sbatch $experiment_id)
 echo -e "Initialization job_id: $init_id\n"
 
@@ -21,7 +22,6 @@ dependency="afterok:$init_id"
 for i in $(seq 0 $((n_iterations - 1)))
 do
     echo "Scheduling iteration $i"
-
     ensemble_array_id=$(
         sbatch --dependency=$dependency --kill-on-invalid-dep=yes --parsable \
                 --job=model-$i \

slurm/initialize.sbatch

@@ -8,7 +8,15 @@ experiment_id=$1
 JULIA_NUM_PRECOMPILE_TASKS=8
 
 echo "Initializing calibration for experiment: $experiment_id"
-julia --color=no --project=experiments/$experiment_id -e 'using Pkg; Pkg.instantiate(;verbose=true)'
+
+julia --color=no --project=experiments/$experiment_id -e '
+    using Pkg; Pkg.instantiate(;verbose=true)
+'
+
+if [ "$generate_data" = true ] ; then
+    echo "Generating observations"
+    julia --project=experiments/$experiment_id experiments/$experiment_id/generate_truth.jl
+fi
 
 julia --color=no --project=experiments/$experiment_id -e '
     import CalibrateAtmos

slurm/model_run.sbatch

@@ -2,19 +2,24 @@
 
 # Extract command-line arguments
 experiment_id=$1
-iteration=$2
+i=$2
 
 # Find output directory
-format_i=$(printf "iteration_%03d" "$iteration")
+format_i=$(printf "iteration_%03d" "$i")
 member=$(printf "member_%03d" "$SLURM_ARRAY_TASK_ID")
-output=output/$experiment_id/$format_i/$member/model_log.out
+output=output/$experiment_id/$format_i/$member/model_log.txt
 
 # Run the forward model
-srun --output=$output julia --color=no --project=experiments/$experiment_id -e "
-    import CalibrateAtmos
-    include(\"experiments/$experiment_id/model_interface.jl\")
+srun --output=$output julia --color=no --project=experiments/$experiment_id -e '
+    import CalibrateAtmos as CAL
+    
+    experiment_id = "'$experiment_id'"
+    i = '$i'
+    member = '$SLURM_ARRAY_TASK_ID'
 
-    physical_model = CalibrateAtmos.get_forward_model(Val(:$experiment_id))
-    config = CalibrateAtmos.get_config(physical_model, $SLURM_ARRAY_TASK_ID, $iteration, \"$experiment_id\")
-    CalibrateAtmos.run_forward_model(physical_model, config)
-"
+    include("experiments/$experiment_id/model_interface.jl")
+
+    physical_model = CAL.get_forward_model(Val(Symbol(experiment_id)))
+    config = CAL.get_config(physical_model, member, i, experiment_id)
+    CAL.run_forward_model(physical_model, config)
+'

slurm/parse_commandline.sh

@@ -3,7 +3,7 @@ slurm_time="2:00:00"
 slurm_ntasks="1"
 slurm_cpus_per_task="1"
 slurm_gpus_per_task="0"
-
+generate_data=false
 help_message="Usage:
     ./pipeline.sh [options] experiment_id
 
@@ -12,13 +12,14 @@ Options:
     -n, --ntasks:        Set number of tasks to launch (default: 1).
     -c, --cpus_per_task: Set CPU cores per task (mutually exclusive with -g, default: 8).
     -g, --gpus_per_task: Set GPUs per task (mutually exclusive with -c, default: 0).
+    --generate_data:     If set, generates observational data for use in the calibration.
     -h, --help:          Display this help message.
 
 Arguments:
     experiment_id:   A unique identifier for your experiment (required)."
 
 # Parse arguments using getopt
-VALID_ARGS=$(getopt -o h,t:,n:,c:,g: --long help,time:,ntasks:,cpus_per_task:,gpus_per_task: -- "$@")
+VALID_ARGS=$(getopt -o h,t:,n:,c:,g: --long help,time:,ntasks:,cpus_per_task:,gpus_per_task:,generate_data -- "$@")
 if [[ $? -ne 0 ]]; then
     exit 1;
 fi
@@ -44,6 +45,10 @@ while [ : ]; do
         slurm_gpus_per_task="$2"
         shift 2
         ;;
+    --generate_data)
+        generate_data=true
+        shift 1
+        ;;
     -h | --help)
         printf "%s\n" "$help_message"
         exit 0
@@ -62,7 +67,7 @@ fi
 ensemble_size=$(grep "ensemble_size:" experiments/$experiment_id/ekp_config.yml | awk '{print $2}')
 n_iterations=$(grep "n_iterations:" experiments/$experiment_id/ekp_config.yml | awk '{print $2}')
 output=$(grep "output_dir:" experiments/$experiment_id/ekp_config.yml | awk '{print $2}')
-logfile=$output/experiment_log.out
+logfile=$output/experiment_log.txt
 
 # Set partition
 if [[ $slurm_gpus_per_task -gt 0 ]]; then

slurm/update.sbatch

@@ -6,17 +6,13 @@ i=$2
 echo "Running update step after iteration $i"
 
 julia --color=no --project=experiments/$experiment_id -e '
-    import YAML, JLD2
-    import CalibrateAtmos
+    import CalibrateAtmos as CAL
 
     experiment_id = "'$experiment_id'"
     i = '$i'
     include("experiments/'$experiment_id'/model_interface.jl")
-    G_ensemble = CalibrateAtmos.observation_map(Val(Symbol(experiment_id)), i)
-    config = YAML.load_file(joinpath("experiments", experiment_id, "ekp_config.yml"))
-    output_dir = config["output_dir"]
-    iter_path = CalibrateAtmos.path_to_iteration(output_dir, i)
-    JLD2.save_object(joinpath(iter_path, "observation_map.jld2"), G_ensemble)
-    CalibrateAtmos.update_ensemble(experiment_id, i)
+    G_ensemble = CAL.observation_map(Val(Symbol(experiment_id)), i)
+    CAL.save_G_ensemble(experiment_id, i, G_ensemble)
+    CAL.update_ensemble(experiment_id, i)
 '
 echo "Update step for iteration $i complete"

src/ekp_interface.jl

@@ -33,6 +33,25 @@ function get_prior(param_dict::AbstractDict; names = nothing)
     return prior
 end
 
+"""
+    get_ekp_config(experiment_id)
+
+Load the EKP configuration for a given `experiment_id`
+"""
+get_ekp_config(experiment_id) =
+    YAML.load_file(joinpath("experiments", experiment_id, "ekp_config.yml"))
+
+"""
+    save_G_ensemble(experiment_id, iteration, G_ensemble)
+
+Save an ensemble's observation map output to the correct folder.
+"""
+function save_G_ensemble(experiment_id, iteration, G_ensemble)
+    config = get_ekp_config(experiment_id)
+    iter_path = path_to_iteration(config["output_dir"], iteration)
+    JLD2.save_object(joinpath(iter_path, "G_ensemble.jld2"), G_ensemble)
+end
+
 """
     initialize(
         experiment_id;
@@ -109,7 +128,7 @@ function update_ensemble(
     eki = JLD2.load_object(eki_path)
 
     # Load data from the ensemble
-    G_ens = JLD2.load_object(joinpath(iter_path, "observation_map.jld2"))
+    G_ens = JLD2.load_object(joinpath(iter_path, "G_ensemble.jld2"))
 
     # Update
     EKP.update_ensemble!(eki, G_ens)
@@ -154,8 +173,7 @@ eki = CalibrateAtmos.calibrate(experiment_id)
 ```
 """
 function calibrate(experiment_id; device = ClimaComms.device())
-    ekp_config =
-        YAML.load_file(joinpath("experiments", experiment_id, "ekp_config.yml"))
+    ekp_config = get_ekp_config(experiment_id)
     # initialize the CalibrateAtmos
     initialize(experiment_id)
 
@@ -179,10 +197,7 @@ function calibrate(experiment_id; device = ClimaComms.device())
 
         # update EKP with the ensemble output and update calibrated parameters
         G_ensemble = observation_map(Val(Symbol(experiment_id)), i)
-        JLD2.save_object(
-            joinpath(path_to_iteration(output_dir, i), "observation_map.jld2"),
-            G_ensemble,
-        )
+        save_G_ensemble(experiment_id, i, G_ensemble)
         eki = update_ensemble(experiment_id, i)
     end
     return eki