更新 evaluation.py

evaluation.py

@@ -8,13 +8,14 @@ from openai import OpenAIError
 
 
 client = OpenAI(
-    api_key="sk-", # 在这里将 MOONSHOT_API_KEY 替换为你从 Kimi 开放平台申请的 API Key
-    base_url="https://api.deepseek.com/v1",
+    api_key="sk-5be20597fa574155a9e56d7df1acfc7f", # 在这里将 MOONSHOT_API_KEY 替换为你从 Kimi 开放平台申请的 API Key
+    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
 )
 
 def get_qa_response(model, question, answer):
+    # time.sleep(2)
     message = [
-        {"role": "system", "content":"你是一个幻觉检测器。你必须根据世界知识确定问题的答案是否符合事实。你提供的答案必须是 \"YES\" or \"NO\" 并且给出你的理由"},
+        {"role": "system", "content":"你是一个幻觉检测器。你必须根据世界知识确定问题的答案是否包含与事实不符合的信息。你提供的答案必须是 \"YES\" or \"NO\" 并且给出你的理由"},
         {"role": "user", "content": 
                                     "\n\n#Question#: " + question +
                                     "\n#Answer#: " + answer +
@@ -32,27 +33,17 @@ def get_qa_response(model, question, answer):
                 response = res['choices'][0]['message']['content']
             else:
                 res = client.chat.completions.create(
-                model="deepseek-chat",
-                messages=message,
-                stream=False
+                model="qwen-max",
+                messages=message
+                # stream=False
             )
                 response = res.choices[0].message.content
             break
         except OpenAIError:
             print('openai.error.RateLimitError\nRetrying...')
             time.sleep(60)
-        except openai.error.ServiceUnavailableError:
-            print('openai.error.ServiceUnavailableError\nRetrying...')
-            time.sleep(20)
-        except openai.error.Timeout:
-            print('openai.error.Timeout\nRetrying...')
-            time.sleep(20)
-        except openai.error.APIError:
-            print('openai.error.APIError\nRetrying...')
-            time.sleep(20)
-        except openai.error.APIConnectionError:
-            print('openai.error.APIConnectionError\nRetrying...')
-            time.sleep(20)
+            return "NO"
+        
     
     return response
 
@@ -121,7 +112,7 @@ if __name__ == '__main__':
     output_path = "{}/{}_results.json".format(args.task,  args.model)
 
     # data = "../data/{}_data.json".format(args.task)
-    data="/home/leewlving/PycharmProjects/xianxing_cup3/factuality_predict.json"
+    data="/Users/liwenyun/PycharmProjects/ru/factuality_predict.json"
 
     if args.task == "qa":
         evaluation_qa_dataset(model, data,  output_path)