Update GitNore

examples/ags/benchmark/humaneval.py

@@ -28,8 +28,9 @@ async def sample_generate(id, result_path:str="samples.jsonl",mode:str="ags"):
         solution_result = await solver.alpha_codium(case['task_id'], case['prompt'], ensemble_count=5)
         sample_dict = dict(task_id=case['task_id'], solution=solution_result['final_solution'])
     elif mode == "llm":
-        solution_result =  await generate_code_block(case['prompt'])
+        solution_result =  await generate_code_block(case['prompt'],case['entry_point'])
         sample_dict = dict(task_id=case['task_id'], solution=solution_result['code_solution'])
+        print(sample_dict)
     with open(result_path, mode='a') as f:
         f.write(json.dumps(sample_dict) + '\n')
     jsonl_ranker(result_path, result_path)

examples/ags/w_action_node/graph.py

@@ -62,6 +62,7 @@ class HumanEvalGraph(Graph):
                 except Exception as e:
                     print(e)
         solution = await self.mdensemble("code", solution_list, problem)
+        print("here",solution)
         solution = await self.tester(problem, rephrase_problem, solution, test_cases)
         return solution
 

examples/ags/w_action_node/operator.py

@@ -332,6 +332,8 @@ class Test(Operator):
             except AssertionError as e:
                 fail_case.append(self.test_cases_2_assert(test_case))
             except Exception as e:
+                with open("tester.txt", "a") as f:
+                    f.write(test_case[0] + "\n")
                 print(e)
                 return {"error":e}
         if fail_case != []:

examples/ags/w_action_node/prompt.py

@@ -55,9 +55,22 @@ You are given a code contest problem, and a self-reflection on the problem:
 The above is an incomplete Python code fragment and reflection on it. Return the complete and correct code with no additional text.
 """
 
-GENERATE_CODEBLOCK_PROMPT = """
-Please provide a self-contained Python script that solves the following problem in a markdown code block:
+# GENERATE_CODEBLOCK_PROMPT = """
+# Please provide a self-contained Python script that solves the following problem in a markdown code block:
+# {problem_description}
+# """
+
+GENERATE_CODEBLOCK_PROMPT ="""
+Please provide a self-contained  Python script that solves the following problem in a markdown code block:
+
 {problem_description}
+
+When creating your solution:
+1. Consider all edge cases and boundary conditions.
+2. Consider the order of operations in your solution and how each step affects subsequent steps.
+3. Avoid oversimplification - address all aspects of the problem.
+4. Ensure your logic covers all stated requirements.
+5. Avoid adding additional test cases beyond those provided in the problem description.
 """
 
 REVIEW_PROMPT = """

examples/ags/w_action_node/utils.py

@@ -159,6 +159,7 @@ async def llm_extract_test_case(id, problem_description: str, file_path:str="pub
 import json
 
 def test_cases_2_test_functions(solution: str, test_case: List):
+    print("here",solution)
     function_name = test_case[0]
     
     def format_param(param):
@@ -181,6 +182,7 @@ def test_cases_2_test_functions(solution: str, test_case: List):
     print(type(test_case[2]), test_case[2])
     expected_output = format_param(test_case[2])
     print(expected_output)
+
     
     tester_function = f"""
 {solution}