Add generation arguments

app.py

@@ -30,6 +30,15 @@ llm = HuggingFaceEndpoint(
 
 client = InferenceClient(endpoint_url, token=token)
 
+gen_kwargs = dict(
+    max_new_tokens=128,
+    top_k=30,
+    top_p=0.95,
+    temperature=0.01,
+    repetition_penalty=1.02,
+    stop_sequences=["\nUser:", "<|endoftext|>", "</s>", "\nHuman:"],
+)
+
 def chat_template_prompt():
     template = """
     Do not repeat questions and do not generate answer for user/human.Do not repeat yourself and do not create/generate dialogues.