reduce repetition penalty

src/cli_demo.py

@@ -53,7 +53,7 @@ def main():
             "temperature": 0.95,
             "num_beams": 1,
             "max_new_tokens": 256,
-            "repetition_penalty": 1.5,
+            "repetition_penalty": 1.0,
             "logits_processor": get_logits_processor()
         }
         with torch.no_grad():

src/web_demo.py

@@ -105,7 +105,7 @@ def predict(input, chatbot, max_length, top_p, temperature, history):
         "temperature": temperature,
         "num_beams": 1,
         "max_length": max_length,
-        "repetition_penalty": 1.5,
+        "repetition_penalty": 1.0,
         "logits_processor": get_logits_processor()
     }
     with torch.no_grad():