Update app.py

app.py

@@ -145,9 +145,9 @@ def chat_response(
     chat_display: str,
     system_prompt: str,
     temperature: float = 0.2,
-    max_new_tokens: int = 8192,
-    top_p: float = 1.0,
-    top_k: int = 20,
+    max_new_tokens: int = 4000,
+    top_p: float = 0.8,
+    top_k: int = 40,
     penalty: float = 1.2,
 ):
     """Generate chat responses, keeping tags visible in the output"""
@@ -249,28 +249,28 @@ def main():
                         minimum=0,
                         maximum=1,
                         step=0.1,
-                        value=1.0,
+                        value=0.2,
                         label="Temperature",
                     )
                     max_tokens = gr.Slider(
                         minimum=128,
                         maximum=32000,
                         step=128,
-                        value=8192,
+                        value=4000,
                         label="Max Tokens",
                     )
                     top_p = gr.Slider(
                         minimum=0.1,
                         maximum=1.0,
                         step=0.1,
-                        value=1.0,
+                        value=0.8,
                         label="Top-p",
                     )
                     top_k = gr.Slider(
                         minimum=1,
                         maximum=100,
                         step=1,
-                        value=20,
+                        value=40,
                         label="Top-k",
                     )
                     penalty = gr.Slider(