fix max tokens (#4)

- update (6920ca85c24d578b94391edce65333ec27208169)
- update (24bc7244ba1c7389b4bcb64a8ce757cd1eebca31)

app.py

@@ -63,7 +63,7 @@ all_input = st.text_area('模型输入', value=prompt, height=400)
 
 
 with st.expander("配置"):
-    stop_tokens = []
+    stop_tokens = ""
     def cut_message(answer: str):
         end = []
         for etk in stop_tokens:
@@ -81,17 +81,17 @@ with st.expander("配置"):
     elif task_type == 'COMPLETION':
         default_top_p, default_temperature, default_n, default_tokens = 0.95, 0.85, 1, 150
     else:
-        default_top_p, default_temperature, default_n, default_tokens = 0.95, 0.85, 5, 85
+        default_top_p, default_temperature, default_n, default_tokens = 0.95, 0.85, 5, 64
     
     model = st.selectbox("model", ["medium", "large", "xl"], index=2)
     top_p = st.slider('top p', 0.0, 1.0, default_top_p)
     top_k = st.slider('top k', 0, 100, 0)
     temperature = st.slider('temperature', 0.0, 1.0, default_temperature)
     n = st.slider('n', 1, 30, default_n)
-    max_tokens = st.slider('max tokens', 4, 1500, default_tokens)
+    max_tokens = st.slider('max tokens', 4, 512, default_tokens)
     
     if st.checkbox("使用换行符作为截断", value=True):
-        stop_tokens.append("\n")
+        stop_tokens = "\n"
 
 def completion():
     resp = requests.post("https://welm.weixin.qq.com/v1/completions", json={
@@ -101,7 +101,8 @@ def completion():
         'top_p': top_p,
         'top_k': top_k,
         'n': n,
-        'model': model
+        'model': model,
+        "stop": stop_tokens,
     }, headers={"Authorization": f"Bearer {st.secrets['token']}"})
     answer = resp.json()
     for idx, choice in enumerate(answer['choices']):