step1: code refactoring: split a large config file to multiple small pieces

step1_config.toml

+model = "/path/to/model"
+apps = "/path/to/apps_dataset"
+
+[sample]
+sample_prompt_path = "path"
+sample_result_path = "path"
+
+[sample.sampling_params]
+n = 0
+temperature = 0.6
+max_new_tokens = 2048
+
+[evaluate]
+evaluate_result_path = ""
+
+[dataset]
+train_path = ""
+test_path = ""
+minimal_test_path = ""
\ No newline at end of file

step1_evaluate_code.py

@@ -147,12 +147,3 @@ def evaluate(code_sample_path, dataset_path, output_path):
 
     results = evaluate_incorrect_code_samples_again(results, apps, 10)
     save_jsonl(results, output_path)
-
-
-if __name__ == "__main__":
-    cfg = read_config()
-    evaluate(
-        cfg["sample"]["sample_result_path"],
-        cfg["apps"],
-        cfg["evaluate"]["evaluate_result_path"],
-    )

step1_run.py

+from utils import read_config
+from utils_vllm import vllm_chatcomplete
+from step1_sample_apps import mk_sample_prompt
+from step1_evaluate_code import evaluate
+from step1_sort_split_dataset import sort_and_split_dataset 
+
+
+if __name__ == "__main__":
+    cfg = read_config()
+    
+    mk_sample_prompt(cfg["model"], cfg["apps"], cfg["sample"]["sample_prompt_path"])
+
+    vllm_chatcomplete(
+        cfg["model"],
+        cfg["sample"]["sample_prompt_path"],
+        cfg["sample"]["sample_result_path"],
+        cfg["sample"]["sampling_params"],
+    )
+
+    evaluate(
+        cfg["sample"]["sample_result_path"],
+        cfg["apps"],
+        cfg["evaluate"]["evaluate_result_path"],
+    )
+
+    sort_and_split_dataset(
+        cfg["evaluate"]["evaluate_result_path"],
+        cfg["dataset"]["train_path"],
+        cfg["dataset"]["test_path"],
+        cfg["dataset"]["minimal_test_path"],
+        cfg["sample"]["sampling_params"]["n"]
+    )
\ No newline at end of file

step1_sample_apps.py

@@ -60,16 +60,3 @@ def mk_sample_prompt(model_path, apps_path, output_path):
 
     print(f"size of dataset: {len(prompts)}")
     save_jsonl(prompts, output_path)
-
-
-if __name__ == "__main__":
-    cfg = read_config()
-    mk_sample_prompt(cfg["model"], cfg["apps"], cfg["sample"]["sample_prompt_path"])
-
-    from utils_vllm import vllm_chatcomplete
-    vllm_chatcomplete(
-        cfg["model"],
-        cfg["sample"]["sample_prompt_path"],
-        cfg["sample"]["sample_result_path"],
-        cfg["sample"]["sampling_params"],
-    )
\ No newline at end of file

step1_sort_split_dataset.py

@@ -56,14 +56,3 @@ def sort_and_split_dataset(raw_dataset_path, new_train_path, new_test_path, mini
     save_jsonl(new_train, new_train_path)
     save_jsonl(new_test, new_test_path)
     save_jsonl(minimal_test, minimal_test_path)
-
-
-if __name__ == "__main__":
-    cfg = read_config()
-    sort_and_split_dataset(
-        cfg["evaluate"]["evaluate_result_path"],
-        cfg["dataset"]["train_path"],
-        cfg["dataset"]["test_path"],
-        cfg["dataset"]["minimal_test_path"],
-        cfg["sample"]["sampling_params"]["n"]
-    )