fix minor optimizers differences

CDB_study.slurm

@@ -7,7 +7,7 @@
 #SBATCH --mem=32G
 #SBATCH --time=48:00:00
 #SBATCH --partition=plgrid-gpu-a100
-#SBATCH -A plgautopt26-gpu-a100
+#SBATCH -A plgrldas2026-gpu-a100
 #SBATCH --array=0-11           # 12 tasks total
 
 SEED=${1:-42}

comprehensive_study.slurm

@@ -7,7 +7,7 @@
 #SBATCH --mem=32G
 #SBATCH --time=48:00:00
 #SBATCH --partition=plgrid-gpu-a100
-#SBATCH -A plgautopt26-gpu-a100
+#SBATCH -A plgrldas2026-gpu-a100
 #SBATCH --array=0-12           # Increased to 13 tasks total to split sequential runs
 
 # 1st argument: SEED (Default: 42)

dynamicalgorithmselection/agents/ppo_utils.py

@@ -159,17 +159,19 @@ def __init__(self, dim, optimizer_num, device):
         super().__init__()
         self.device = device
         self.optimizer_num = optimizer_num
-        self.embedders = [
-            (
-                nn.Sequential(
-                    nn.Linear(dim, 64),
-                    nn.ReLU(),
-                    nn.Linear(64, 1),
-                    nn.ReLU(),
-                )
-            ).to(device)
-            for _ in range(2 * optimizer_num)
-        ]
+        self.embedders = nn.ModuleList(
+            [
+                (
+                    nn.Sequential(
+                        nn.Linear(dim, 64),
+                        nn.ReLU(),
+                        nn.Linear(64, 1),
+                        nn.ReLU(),
+                    )
+                ).to(device)
+                for _ in range(2 * optimizer_num)
+            ]
+        )
 
         self.embedder_final = nn.Sequential(
             nn.Linear(9 + optimizer_num * 2, 64),
@@ -203,17 +205,19 @@ class RLDASCritic(nn.Module):
     def __init__(self, dim, optimizer_num, device):
         super().__init__()
         self.device = device
-        self.embedders = [
-            (
-                nn.Sequential(
-                    nn.Linear(dim, 64),
-                    nn.ReLU(),
-                    nn.Linear(64, 1),
-                    nn.ReLU(),
-                )
-            ).to(device)
-            for _ in range(2 * optimizer_num)
-        ]
+        self.embedders = nn.ModuleList(
+            [
+                (
+                    nn.Sequential(
+                        nn.Linear(dim, 64),
+                        nn.ReLU(),
+                        nn.Linear(64, 1),
+                        nn.ReLU(),
+                    )
+                ).to(device)
+                for _ in range(2 * optimizer_num)
+            ]
+        )
         self.embedder_final = nn.Sequential(
             nn.Linear(9 + optimizer_num * 2, 64),
             nn.Tanh(),

dynamicalgorithmselection/main.py

@@ -1,6 +1,7 @@
 import argparse
 import os
 import shutil
+import warnings
 from random import seed as set_random_seed
 from typing import List, Type, Dict, Any
 import cocopp
@@ -17,6 +18,8 @@
 from dynamicalgorithmselection.experiments.utils import DIMENSIONS
 from dynamicalgorithmselection.optimizers.Optimizer import Optimizer
 
+warnings.filterwarnings("ignore")
+
 AGENTS_DICT = {
     "random": RandomAgent,
     "policy-gradient": PolicyGradientAgent,
@@ -144,7 +147,7 @@ def parse_arguments():
         "--reward-option",
         type=int,
         choices=[1, 2, 3, 4],
-        default=2,
+        default=1,
         help="id of method used to compute reward (valid options: 1-4, default 2 maps to r2)",
     )
 
@@ -201,7 +204,6 @@ def test(args, action_space):
     options = {
         "action_space": action_space,
     } | common_options(args)
-    # agent_state = torch.load(f)
     if args.agent == "policy-gradient":
         options.update(
             torch.load(

dynamicalgorithmselection/optimizers/DE/DE.py

@@ -9,6 +9,8 @@ def __init__(self, problem, options):
             self.n_individuals is None
         ):  # number of offspring, aka offspring population size
             self.n_individuals = 170
+        self.Nmin = 30  # Discrepancy: Nmin is 30 in Population.py
+        self.Nmax = 170
         assert self.n_individuals > 0
         self._n_generations = 0  # number of generations
         self._printed_evaluations = self.n_function_evaluations