Add global seed setting to all benchmarks, rerun results and sort res…

…ults by benchmark name for consistent table generation

benchmarking/generate_summary.py

@@ -37,7 +37,7 @@ def main():
     args = input_arguments()
     results_dict = defaultdict(lambda: defaultdict(pd.DataFrame))
     base_result_path = Path("benchmarking/results")
-    for file_path in base_result_path.glob("**/traj.parquet.gzip"):
+    for file_path in sorted(base_result_path.glob("**/traj.parquet.gzip")):
         version = file_path.parts[-3]
         benchmark = file_path.parts[-2]
 

benchmarking/hpobench_benchmark.py

@@ -1,4 +1,5 @@
 import argparse
+import random
 from pathlib import Path
 
 import numpy as np
@@ -159,6 +160,8 @@ def main():
         trajectories = []
         for seed in seeds:
             print(f"Running benchmark {benchmark_name} on seed {seed}")
+            np.random.seed(seed)
+            random.seed(seed)
             dehb_params["seed"] = int(seed)
             dehb_optimizer = DEHBOptimizerHPOBench(
                 dehb_params=dehb_params,

benchmarking/mfpbench_benchmark.py

@@ -1,4 +1,5 @@
 import argparse
+import random
 from pathlib import Path
 
 import mfpbench
@@ -122,6 +123,8 @@ def main():
         trajectories = []
         for seed in seeds:
             print(f"Running benchmark {benchmark_name} on seed {seed}")
+            np.random.seed(seed)
+            random.seed(seed)
             dehb_params["seed"] = int(seed)
             dehb_optimizer = DEHBOptimizerMFPBench(
                 dehb_params=dehb_params,