update

backend-cli.py

@@ -100,6 +100,9 @@ def process_finished_requests() -> bool:
     # Sort the evals by priority (first submitted first run)
     eval_requests: list[EvalRequest] = sort_models_by_priority(api=API, models=eval_requests)
 
+    # XXX
+    # eval_requests = [r for r in eval_requests if 'bloom-560m' in r.model]
+
     random.shuffle(eval_requests)
 
     from src.leaderboard.read_evals import get_raw_eval_results
@@ -161,7 +164,10 @@ def process_pending_requests() -> bool:
     set_eval_request(api=API, eval_request=eval_request, set_to_status=RUNNING_STATUS, hf_repo=QUEUE_REPO,
                      local_dir=EVAL_REQUESTS_PATH_BACKEND)
 
-    for task in TASKS_HARNESS:
+    task_lst = TASKS_HARNESS.copy()
+    random.shuffle(task_lst)
+
+    for task in task_lst:
         results = process_evaluation(task, eval_request)
 
     set_eval_request(api=API, eval_request=eval_request, set_to_status=FINISHED_STATUS, hf_repo=QUEUE_REPO,
@@ -171,6 +177,7 @@ def process_pending_requests() -> bool:
 
 
 if __name__ == "__main__":
+    # res = False
     res = process_pending_requests()
 
     if res is False: