refactor(sft): 重构 SFT 数据处理脚本

sft_coder.py

@@ -0,0 +1,21 @@
+# inclusionAI/Ling-Coder-SFT
+from datasets import load_dataset
+from utils import process_dataset
+
+
+def process_func(input_dict: dict) -> dict:
+    msg = input_dict["messages"]
+    query = msg[0]["content"]
+    history = msg[1]["content"]
+    return {"query": query, "response": history}
+
+
+if __name__ == "__main__":
+    dataset = load_dataset("inclusionAI/Ling-Coder-SFT")
+    
+    process_dataset(
+        dataset_dict=dataset,
+        output_subdir="Ling-Coder-SFT",
+        process_func=process_func
+    )
+    

utils.py

@@ -102,7 +102,7 @@ def dump_pkl_files(
             sequence = torch.cat(package[key])
             output_package[key] = sequence
          
-        with open(out_file_path, "w") as f:
+        with open(out_file_path, "wb") as f:
             pkl.dump(output_package, f)