Add local run support

yourbench_space/app.py

@@ -144,7 +144,11 @@ def enable_button(files):
 def run_evaluation_pipeline(oauth_token: gr.OAuthToken | None, org_name, eval_name, config_name="lighteval"):
     eval_ds_name = f"{org_name}/{eval_name}"
     repo_id = f"{org_name}/leaderboard_yourbench_{eval_ds_name.replace('/', '_')}"
-    folder_path = str(Path(PATH) / "yourbench_space" / "leaderboard_space")
+
+    if os.environ.get("SYSTEM") == "spaces":
+        folder_path = str(PATH / "yourbench_space" / "leaderboard_space")
+    else:
+        folder_path = str(PATH.parent / "yourbench_space" / "leaderboard_space")
 
     new_env = os.environ.copy()
     if oauth_token:

yourbench_space/evaluation.py

@@ -14,7 +14,12 @@ OUTPUT_DIR = "/data" if ON_SPACES else "."  # TODO: fix the space folder
 
 def create_eval_file(eval_ds_name: str):
     task_name = eval_ds_name.replace("/", "_")
-    template_path = Path("/home/user/app/yourbench_space/lighteval_task/yourbench_task.py")
+
+    if os.environ.get("SYSTEM") == "spaces":
+        template_path = Path("/home/user/app/yourbench_space/lighteval_task/yourbench_task.py")
+    else:
+         template_path = Path("yourbench_space/lighteval_task/yourbench_task.py")
+
     subprocess.run(["lighteval", "tasks", "create", str(template_path), task_name, eval_ds_name])
 
 
@@ -40,7 +45,7 @@ async def run_evaluations(eval_ds_name: str, org: str, custom_env=None) -> list:
             "lighteval",
             "endpoint",
             "inference-providers",
-            f"model={model_name},provider={provider}",
+            f"model_name={model_name},provider={provider}",
             f"custom|{task_name}|0|0",
             "--custom-tasks",
             f"custom_{task_name}_task.py",