[EXP ONLY] Only use us locations for VertexAI

ci/k8s/pr-exp.yaml

@@ -44,11 +44,11 @@ spec:
           name: results-volume
         env:
         - name: LLM_NUM_EXP
-          value: '40'
+          value: '1'
         - name: LLM_NUM_EVA
-          value: '10'
+          value: '1'
         - name: VERTEX_AI_LOCATIONS
-          value: 'asia-east1,asia-east2,asia-northeast1,asia-northeast3,asia-south1,asia-southeast1,australia-southeast1,europe-central2,europe-north1,europe-southwest1,europe-west1,europe-west2,europe-west3,europe-west4,europe-west6,europe-west8,europe-west9,southamerica-east1,us-central1,us-east1,us-east4,us-east5,us-south1,us-west1,us-west4'
+          value: 'us-central1'
         - name: CLOUD_BUILD_LOCATION
           value: 'us-west2'
         - name: GCB_BUILDPOOL_NAME

run_all_experiments.py

@@ -270,7 +270,7 @@ def parse_args() -> argparse.Namespace:
   parser.add_argument('-mr',
                       '--agent-max-round',
                       type=int,
-                      default=10,
+                      default=100,
                       help='Max trial round for agents.')
 
   args = parser.parse_args()

run_one_experiment.py

@@ -292,7 +292,7 @@ def _fuzzing_pipelines(benchmark: Benchmark, model: models.LLM,
   with pool.ThreadPool(processes=NUM_EVA) as p:
     # Initialize thread-local storage in each worker before processing
     task_args = [(benchmark, model, args, work_dirs, trial)
-                 for trial in range(1, NUM_EVA + 1)]
+                 for trial in range(1, 11)]
     trial_results = p.starmap(_fuzzing_pipeline, task_args)
   return BenchmarkResult(benchmark=benchmark,
                          work_dirs=work_dirs,