reporeformat V2 (#23)

* reformat factor implement process

* move some code to more reasonable place

* fix the bug

* add test function in factor_extract_and_implement.py

* change select factor number to ratio , add some factor implement setting and fix some bug while using knowledgebase

* change evoagent

* add abstract class EvoAgent

* add benchmark workflow

* fix some bug in llm_utils

* run wenjun's code

* fix the knowledgebase instance check

---------

Co-authored-by: xuyang1 <xuyang1@microsoft.com>

rdagent/app/factor_implementation/eval_implement.py

@@ -0,0 +1,30 @@
+from rdagent.core.conf import BenchmarkSettings
+from rdagent.core.utils import import_class
+from rdagent.benchmark.eval_method import FactorImplementEval
+from rdagent.benchmark.data_process import load_eval_data
+
+# 1.read the settings
+bs = BenchmarkSettings()
+
+# 2.read and prepare the eval_data
+test_cases = load_eval_data(bs.bench_version)
+
+# 3.declare the method to be tested and pass the arguments.
+# TODO: Whether it is necessary to define two Eval method classes for two data type?
+method_cls = import_class(bs.bench_method_cls)
+generate_method = method_cls(bs.bench_method_extra_kwargs)
+
+# 4.declare the eval method and pass the arguments.
+eval_method = FactorImplementEval(
+    method=generate_method,
+    test_cases=test_cases,
+    catch_eval_except=True,
+    test_round=bs.bench_test_round,
+)
+
+# 5.run the eval
+eval_method.eval()
+
+# 6.save the result
+eval_method.save(output_path = bs.bench_result_path)
+