fix merge problem

pilot/server/sqlgpt.py

@@ -7,7 +7,6 @@ import torch
 import gradio as gr
 from fastchat.serve.inference import generate_stream, compress_module
 
-
 from transformers import AutoTokenizer, AutoModelForCausalLM
 device = "cuda" if torch.cuda.is_available() else "cpu"
 BASE_MODE = "/home/magic/workspace/github/DB-GPT/models/vicuna-13b"
@@ -21,12 +20,12 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 
 def generate(prompt):    
-    # compress_module(model, device) 
-    # model.to(device)
+    compress_module(model, device) 
+    model.to(device)
     print(model, tokenizer)
     params = {
         "model": "vicuna-13b",
-        "prompt": prompt,
+        "prompt": "这是一个用户与助手之间的对话, 助手精通数据库领域的知识, 并能够对数据库领域知识做出非常专业的回答。以下是用户的问题:" + prompt,
         "temperature": 0.7,
         "max_new_tokens": 512,
         "stop": "###"
@@ -36,9 +35,6 @@ def generate(prompt):
         
     for chunk in output:
         yield chunk
-    #for chunk in output.iter_lines(decode_unicode=False, delimiter=b"\0"):
-    #    if chunk:
-    #        yield chunk
 
 if __name__ == "__main__":
     with gr.Blocks() as demo:
@@ -53,5 +49,3 @@ if __name__ == "__main__":
 
     demo.queue(concurrency_count=3).launch(server_name="0.0.0.0") 
 
-
-