Update app.py

app.py

@@ -12,7 +12,7 @@ device = torch.device("cuda:{}".format(device_ids_parallel[0]) if USE_CUDA else
 # 初始化
 peft_model_id = "CMLM/ZhongJing-2-1_8b"
 base_model_id = "Qwen/Qwen1.5-1.8B-Chat"
-model = AutoModelForCausalLM.from_pretrained(base_model_id, device_map="auto")
+model = AutoModelForCausalLM.from_pretrained(base_model_id, device_map="auto").to(device)
 model.load_adapter(peft_model_id)
 tokenizer = AutoTokenizer.from_pretrained(
     "CMLM/ZhongJing-2-1_8b",
@@ -29,8 +29,8 @@ def single_turn_chat(question):
         {"role": "system", "content": "You are a helpful TCM medical assistant named 仲景中医大语言模型, created by 医哲未来 of Fudan University."},
         {"role": "user", "content": prompt}
     ]
-    input = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-    model_inputs = tokenizer([input], return_tensors="pt").to(device)
+    input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    model_inputs = tokenizer([input_text], return_tensors="pt").to(device)
     generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=512)
     generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
     response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
@@ -63,16 +63,13 @@ def multi_turn_chat(question, chat_history=None):
     chat_history.append({"role": "assistant", "content": response})
     
     # Format the chat history for output
-    tempass = ""
-    tempuser = ""
     formatted_history = []
+    tempuser = ""
     for entry in chat_history:
         if entry['role'] == 'user':
             tempuser = entry['content']
         elif entry['role'] == 'assistant':
-            tempass = entry['content']
-            temp = (tempuser, tempass)
-            formatted_history.append(temp)
+            formatted_history.append((tempuser, entry['content']))
     
     return formatted_history, chat_history
 
@@ -102,4 +99,4 @@ with gr.Blocks() as multi_turn_interface:
     user_input.submit(multi_turn_chat, [user_input, state], [chatbot, state])
 
 single_turn_interface.launch()
-multi_turn_interface.launch()
+multi_turn_interface.launch()