update scripts

app.py

@@ -15,6 +15,7 @@ import src.submission.submit as submit
 import os
 TOKEN = os.environ.get("H4_TOKEN", None)
 print("TOKEN", TOKEN)
+import src.backend.run_eval_suite as run_eval_suite
 
 def ui_snapshot_download(repo_id, local_dir, repo_type, tqdm_class, etag_timeout):
     try:
@@ -41,6 +42,33 @@ def init_space():
 original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
 
 
+def process_pending_evals():
+    if len(pending_eval_queue_df) == 0:
+        print("No pending evaluations found.")
+        return
+
+    for _, eval_request in pending_eval_queue_df.iterrows():
+        # 根据具体的列名和数据格式调整 `eval_request` 的内容
+        print(f"Evaluating model: {eval_request['model']}")
+
+        # 调用评估函数
+        run_eval_suite.run_evaluation(
+            eval_request=eval_request,
+            local_dir=envs.EVAL_RESULTS_PATH_BACKEND,
+            results_repo=envs.RESULTS_REPO,
+            batch_size=1,
+            device=envs.DEVICE,
+            no_cache=True,
+            need_check=True,  # 根据需要设定是否需要检查
+            write_results=True  # 根据需要设定是否写入结果
+        )
+        print(f"Finished evaluation for model: {eval_request['model']}")
+
+
+# 在初始化完成后调用
+original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
+process_pending_evals()
+
 # try:
 #     print(envs.EVAL_REQUESTS_PATH)
 #     snapshot_download(