fix bugs

download_model.py

@@ -1,3 +1,3 @@
 from modelscope import snapshot_download
-model_dir = snapshot_download('qwen/Qwen-1_8B-Chat', cache_dir='../download_llm')
+model_dir = snapshot_download('qwen/Qwen1.5-0.5B', cache_dir='../download_llm')
 print("模型下载完成")

llm_tricks/DPO_example/dataset.py

@@ -1,34 +1,38 @@
 import torch
 from torch.utils.data import Dataset
+import json
 
 
 class RlhfDataset(Dataset):
-    def __init__(self, data, tokenizer):
-        self.data = data
-
-        self.encoded_texts = []
-        for entry in data:
-            prompt = entry["prompt"]
-            rejected_response = entry["rejected"]
-            chosen_response = entry["chosen"]
-
-            prompt_tokens = tokenizer.encode(prompt)
-            chosen_full_text = f"{prompt}\n\n### Response:\n{chosen_response}"
-            rejected_full_text = f"{prompt}\n\n### Response:\n{rejected_response}"
-            chosen_full_tokens = tokenizer.encode(chosen_full_text)
-            rejected_full_tokens = tokenizer.encode(rejected_full_text)
-
-            self.encoded_texts.append({
+    def __init__(self, file_path, tokenizer):
+        with open(file_path, "r", encoding="utf-8") as file:
+            data_list = file.readlines()
+        self.data_list = data_list
+        self.tokenizer = tokenizer
+
+    def __getitem__(self, item):
+        data = self.data_list[item]
+        data = json.loads(data)
+        prompt = data['prompt']
+        chosen = data['chosen']
+        rejected = data['rejected']
+
+        chosen_full_text = f"{prompt}\n\n### Response:\n{chosen}"
+        rejected_full_text = f"{prompt}\n\n### Response:\n{rejected}"
+
+        prompt_tokens = self.tokenizer.encode(prompt, add_special_tokens=False)
+        chosen_full_tokens = self.tokenizer.encode(chosen_full_text, add_special_tokens=False)
+        rejected_full_tokens = self.tokenizer.encode(rejected_full_text, add_special_tokens=False)
+
+        input = {
                 "prompt": prompt_tokens,
                 "chosen": chosen_full_tokens,
                 "rejected": rejected_full_tokens,
-            })
-
-    def __getitem__(self, index):
-        return self.encoded_texts[index]
+            }
+        return input
 
     def __len__(self):
-        return len(self.data)
+        return len(self.data_list)
 
 
 def data_collate(