Update app.py

app.py

@@ -18,14 +18,14 @@ model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float
 
 # 생성 설정 (Gradio UI에서 제어할 수 있는 변수들)
 default_generation_config = GenerationConfig(
-    temperature=0.7,
-    top_k=50,
-    top_p=0.95,
+    temperature=0.1,
+    top_k=30,
+    top_p=0.5,
     do_sample=True,
     num_beams=1,
     repetition_penalty=1.1,
     min_new_tokens=10,
-    max_new_tokens=512
+    max_new_tokens=30
 )
 
 # 응답 생성 함수
@@ -47,7 +47,12 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
     
     # 모델 입력 생성
     inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to(model.device)
-    response_ids = model.generate(**inputs, generation_config=generation_config)
+    response_ids = model.generate(
+        **inputs, 
+        generation_config=generation_config, 
+        eos_token_id=tokenizer.eos_token_id,  # 종료 토큰 설정
+        pad_token_id=tokenizer.eos_token_id   # pad_token_id도 종료 토큰으로 설정
+    )
     
     # 모델 응답 디코딩
     response_text = tokenizer.decode(response_ids[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
@@ -64,9 +69,9 @@ demo = gr.ChatInterface(
     respond,
     additional_inputs=[
         gr.Textbox(value="You are a friendly Chatbot that recommends music.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
+        gr.Slider(minimum=1, maximum=2048, value=30, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.1, step=0.1, label="Temperature"),
+        gr.Slider(minimum=0.1, maximum=1.0, value=0.5, step=0.05, label="Top-p (nucleus sampling)"),
     ],
 )