From e6787a2036e936dc1c09b909bfbb3fa121628151 Mon Sep 17 00:00:00 2001
From: ViperEkura <3081035982@qq.com>
Date: Fri, 17 Jul 2026 23:15:18 +0800
Subject: [PATCH] feat: add BELLE SFT export script, remove
 Firefly-1.1M-Rephrased

---
 .../{sft_firefly-1.1m-rephrased.py => sft_belle.py}  | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)
 rename scripts/supervised_finetuning/{sft_firefly-1.1m-rephrased.py => sft_belle.py} (53%)

diff --git a/scripts/supervised_finetuning/sft_firefly-1.1m-rephrased.py b/scripts/supervised_finetuning/sft_belle.py
similarity index 53%
rename from scripts/supervised_finetuning/sft_firefly-1.1m-rephrased.py
rename to scripts/supervised_finetuning/sft_belle.py
index 76d6e90..f92dac5 100644
--- a/scripts/supervised_finetuning/sft_firefly-1.1m-rephrased.py
+++ b/scripts/supervised_finetuning/sft_belle.py
@@ -3,17 +3,23 @@ from pipeline import export_dataset
 
 
 def process_func(input_dict: dict):
+    instruction = input_dict["instruction"]
+    inp = input_dict.get("input", "")
+    if inp:
+        content = instruction + "\n" + inp
+    else:
+        content = instruction
     return {"messages": [
-        {"role": "user", "content": input_dict["instruction"]},
+        {"role": "user", "content": content},
         {"role": "assistant", "content": input_dict["output"]},
     ]}
 
 
 if __name__ == "__main__":
-    dataset = load_dataset("Mxode/Firefly-1.1M-Rephrased")
+    dataset = load_dataset("BelleGroup/train_2M_CN")
     export_dataset(
         dataset=dataset["train"],
         output_dir="./dataset",
-        output_prefix="Firefly-1.1M-Rephrased",
+        output_prefix="belle-sft",
         process_func=process_func,
     )