increase # of MAX_TOKEN & remove the sample reasoning/answer box

app.py

@@ -9,14 +9,18 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStream
 
 os.environ.setdefault("TOKENIZERS_PARALLELISM", "false")
 
+# 두 가지 수정을 해야 함
+# 1. Sample reasoning과 Sample answer box 지우기
+# 2. MAX_TOKEN 늘려서 끝까지 생성하고 reasoning과 assistant가 모두 생성된 스크린샷 찍기
+# 3. (Optional) system prompt / your message도 수정해서 새로운 prompt-message 조합으로 실행
 MODEL_ID = os.getenv("MODEL_ID", "Qwen/Qwen3-0.6B")
-MAX_NEW_TOKENS = int(os.getenv("MAX_NEW_TOKENS", "256"))
+MAX_NEW_TOKENS = int(os.getenv("MAX_NEW_TOKENS", "2048")) # 256
 MAX_INPUT_TOKENS = int(os.getenv("MAX_INPUT_TOKENS", "1536"))
 MAX_HISTORY_TURNS = int(os.getenv("MAX_HISTORY_TURNS", "3"))
 N_THREADS = int(os.getenv("N_THREADS", str(max(1, os.cpu_count() or 1))))
 DEFAULT_SYSTEM_PROMPT = os.getenv(
     "SYSTEM_PROMPT",
-    "You are a helpful assistant. Keep answers clear and concise.",
+    "You are a helpful assistant. Keep answers clear and concise. If user",
 )
 
 PRESETS = {
@@ -295,19 +299,19 @@ with gr.Blocks(title="Local CPU split-reasoning chat") as demo:
         lines=4,
     )
 
-    with gr.Row():
-        sample_reasoning = gr.Textbox(
-            label="Sample reasoning",
-            value=PRESETS["Math"]["sample_reasoning"],
-            lines=5,
-            interactive=False,
-        )
-        sample_answer = gr.Textbox(
-            label="Sample answer",
-            value=PRESETS["Math"]["sample_answer"],
-            lines=5,
-            interactive=False,
-        )
+    # with gr.Row():
+    #     sample_reasoning = gr.Textbox(
+    #         label="Sample reasoning",
+    #         value=PRESETS["Math"]["sample_reasoning"],
+    #         lines=5,
+    #         interactive=False,
+    #     )
+    #     sample_answer = gr.Textbox(
+    #         label="Sample answer",
+    #         value=PRESETS["Math"]["sample_answer"],
+    #         lines=5,
+    #         interactive=False,
+    #     )
 
     with gr.Row():
         send_btn = gr.Button("Send", variant="primary")