updated bash script

Author: codegithubka

docs/HPC_GUIDE.md

@@ -39,7 +39,7 @@ scancel <JOBID>
 
 # Monitoring live progress
 
-tail -f logs_<JOBID>.err
+tail -f logs_18702594.err
 
 # Watch task completetion
 

hpc/run_phase3.sh

@@ -1,9 +1,9 @@
 #!/bin/bash
 #SBATCH --job-name=pp_phase3
-#SBATCH --partition=rome
+#SBATCH --partition=genoa
 #SBATCH --nodes=1
 #SBATCH --ntasks=1
-#SBATCH --cpus-per-task=32
+#SBATCH --cpus-per-task=64
 #SBATCH --time=02:00:00
 #SBATCH --mem=64G
 #SBATCH --output=/home/kanagnostopoul/CSS_Project/pp_phase3_%j.out
@@ -35,7 +35,7 @@ echo "CPUs:       $SLURM_CPUS_PER_TASK"
 echo "Start:      $(date)"
 echo "Working dir: $(pwd)"
 echo "========================================"
-
+x
 # -----------------------------------------------------------------------------
 # Environment Setup
 # -----------------------------------------------------------------------------

hpc/run_phase4.sh

@@ -4,8 +4,8 @@
 #SBATCH --nodes=1
 #SBATCH --ntasks=1
 #SBATCH --cpus-per-task=64
-#SBATCH --time=04:00:00
-#SBATCH --mem=64G
+#SBATCH --time=08:00:00
+#SBATCH --mem=16G
 #SBATCH --output=/home/kanagnostopoul/CSS_Project/pp_phase4_%j.out
 #SBATCH --error=/home/kanagnostopoul/CSS_Project/pp_phase4_%j.err
 

models/config.py

@@ -238,33 +238,17 @@ def estimate_runtime(self, n_cores: int = 32) -> str:
 
 # Phase 4: Sensitivity analysis
 PHASE4_CONFIG = Config(
-    grid_size=250,
-    predator_birth_values=(0.15, 0.20, 0.25, 0.30),
-    predator_death_values=(0.05, 0.10, 0.15, 0.20),
-    n_prey_death=10,
-    prey_death_range=(0.01, 0.10),
-    n_replicates=10,
-    warmup_steps=500,
-    measurement_steps=500,
+    grid_size=250,          # As requested
+    n_replicates=10,        # As requested
+    warmup_steps=500,       # As requested
+    measurement_steps=500,  # As requested
     with_evolution=False,
     collect_pcf=False,
     save_timeseries=True,
     timeseries_subsample=10,
-    
-   
 )
-"""
-Phase 4 Reconfiguration Notes:
 
-We vary evetyhing
-pred death and birth
-prey death and birth
-11 values for each from 0 to 1
-10 reps
-grid size of 250
-64 cores
-500 + 500 warmup and measuremnt
-"""
+
 # Phase 5: Perturbation analysis (critical slowing down)
 PHASE5_CONFIG = Config(
     grid_size=100,

scripts/experiments.py

@@ -508,60 +508,79 @@ def run_phase3(cfg: Config, output_dir: Path, logger: logging.Logger) -> List[Di
 
 def run_phase4(cfg: Config, output_dir: Path, logger: logging.Logger) -> List[Dict]:
     """
-    Phase 4: Sensitivity analysis.
-    - Vary predator_birth and predator_death
-    - For each combo, sweep prey_death
-    
-    
-    NOTE: This phase should be subjected to changes depeding on what we are interested in varying
-    in terms of parameters.
+    Phase 4: Global Sensitivity Analysis.
+    Vary: prey_birth, prey_death, predator_birth, predator_death
+    Range: 0-1 (11 values each)
+    Reps: 10
+    Grid size: 250
     """
     from joblib import Parallel, delayed
+    import itertools
 
     warmup_numba_kernels(cfg.grid_size, directed_hunting=cfg.directed_hunting)
 
-    prey_deaths = np.linspace(cfg.prey_death_range[0], cfg.prey_death_range[1], cfg.n_prey_death)
+    # Define the global sweep values
+    sweep_values = np.linspace(0.0, 1.0, 11)
+    
+    # Logging
+    logger.info(f"Phase 4: Full 4D Parameter Sweep")
+    logger.info(f"  Parameters: prey_birth, prey_death, pred_birth, pred_death")
+    logger.info(f"  Range: 0.0 to 1.0 (11 steps)")
+    logger.info(f"  Grid Size: {cfg.grid_size}")
+    logger.info(f"  Replicates: {cfg.n_replicates}")
+    
+    param_grid = itertools.product(sweep_values, repeat=4)
 
     jobs = []
-    for pred_b in cfg.predator_birth_values:
-        for pred_d in cfg.predator_death_values:
-            for pd in prey_deaths:
-                for rep in range(cfg.n_replicates):
-                    params = {"pred_b": pred_b, "pred_d": pred_d, "pd": pd}
-                    
-                    # Non-evolution
-                    seed = generate_unique_seed(params, rep)
-                    jobs.append((cfg.prey_birth, pd, pred_b, pred_d,
-                                cfg.grid_size, seed, cfg, False))
-                    
-                    # Evolution
-                    evo_seed = generate_unique_seed(params, rep + 1_000_000)
-                    jobs.append((cfg.prey_birth, pd, pred_b, pred_d,
-                                cfg.grid_size, evo_seed, cfg, True))
-    
-    logger.info(f"Phase 4: {len(jobs):,} simulations")
-    logger.info(f"  predator_birth: {cfg.predator_birth_values}")
-    logger.info(f"  predator_death: {cfg.predator_death_values}")
 
+    for pb, pd, pred_b, pred_d in param_grid:
+        for rep in range(cfg.n_replicates):
+            # Create params dict for unique seed generation
+            params_id = {
+                "pb": pb, 
+                "pd": pd, 
+                "pred_b": pred_b, 
+                "pred_d": pred_d, 
+                "rep": rep
+            }
+            seed = generate_unique_seed(params_id, rep)
+            
+            # Job Signature: 
+            # (prey_birth, prey_death, predator_birth, predator_death, grid_size, seed, cfg, with_evolution)
+            jobs.append((
+                pb, 
+                pd, 
+                pred_b, 
+                pred_d,
+                cfg.grid_size, 
+                seed, 
+                cfg, 
+                False 
+            ))
+    
+    logger.info(f"  Total simulations: {len(jobs):,}")
     output_jsonl = output_dir / "phase4_results.jsonl"
     all_results = []
 
     with open(output_jsonl, "w", encoding="utf-8") as f:
         executor = Parallel(n_jobs=cfg.n_jobs, return_as="generator")
         tasks = (delayed(run_single_simulation)(*job) for job in jobs)
 
-        for result in tqdm(executor(tasks), total=len(jobs), desc="Phase 4"):
+        # tqdm shows progress for the massive job list
+        for result in tqdm(executor(tasks), total=len(jobs), desc="Phase 4 (4D Sweep)"):
             f.write(json.dumps(result, default=str) + "\n")
             f.flush()
             all_results.append(result)
 
+    # 4. Save Metadata
     meta = {
         "phase": 4,
-        "description": "Sensitivity analysis",
-        "predator_birth_values": list(cfg.predator_birth_values),
-        "predator_death_values": list(cfg.predator_death_values),
+        "description": "Global 4D Sensitivity Analysis",
+        "sweep_values": sweep_values.tolist(),
+        "parameters_varied": ["prey_birth", "prey_death", "predator_birth", "predator_death"],
         "n_sims": len(all_results),
         "timestamp": time.strftime("%Y-%m-%d %H:%M:%S"),
+        "config": asdict(cfg),
     }
     with open(output_dir / "phase4_metadata.json", "w") as f:
         json.dump(meta, f, indent=2, default=str)