新增了参数评估脚本和评分脚本

backend/test1/evaluator.py

@@ -14,14 +14,20 @@ from llama_index.core import (
     SimpleDirectoryReader,
     Response,
 )
-from llama_index.core.evaluation import FaithfulnessEvaluator, DatasetGenerator
+from llama_index.core.evaluation import (
+    FaithfulnessEvaluator, 
+    DatasetGenerator, 
+    CorrectnessEvaluator, 
+    SemanticSimilarityEvaluator,)
 
 
 
 init_settings()
 init_observability()
 
-evaluator_qwen = FaithfulnessEvaluator()
+faith_evaluator_qwen = FaithfulnessEvaluator() #诚实度评测
+corr_evaluator_qwen = CorrectnessEvaluator() #准确率评测
+Seman_evaluator_qwen = SemanticSimilarityEvaluator()#嵌入相似度评估
 
 documents = SimpleDirectoryReader("D:/LLM_model/text2sql/zjdataai-app-test/backend/data-test").load_data()
 
@@ -52,12 +58,12 @@ async def evaluate_query_engine_async(query_engine, questions):
     c = [query_engine.aquery(q) for q in questions]
     gathering_future = asyncio.gather(*c)
     results = await gathering_future
-    print(results)
+    #print(results)
     
     total_correct = 0
     for r in results:
         eval_result = (
-            1 if evaluator_qwen.evaluate_response(response=r).passing else 0
+            1 if faith_evaluator_qwen.evaluate_response(response=r).passing else 0
         )
         total_correct += eval_result