Update app.py

app.py

@@ -1,42 +1,39 @@
 import gradio as gr
 import time
-import psutil
 from optimum.intel import OVModelForCausalLM
 from transformers import AutoTokenizer, pipeline
 
-# 載入模型和 tokenizer
+# Load the model and tokenizer
 model_id = "hsuwill000/DeepSeek-R1-Distill-Qwen-1.5B-openvino"
-model = OVModelForCausalLM.from_pretrained(model_id, device="CPU")  # 明確指定設備
+model = OVModelForCausalLM.from_pretrained(model_id, device="CPU")  # 明确指定设备
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 
-# 建立生成 pipeline
+# Create generation pipeline
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
 def respond(message):
-    # 取得 CPU 使用率（0.1 秒內的平均值）
-    cpu_usage = psutil.cpu_percent(interval=0.1)
-    if cpu_usage > 80:
-        # CPU 使用率超過 50%，直接返回忙碌提示訊息
-        return [(message, "系統目前忙碌中，請稍候...")]
-    else:
+    try:
         start_time = time.time()
-        # 強化 prompt，要求回答簡明且不重複
+        
+        # 強化 Prompt 讓模型輸出更合理
         instruction = (
             "請用簡單、準確的語言回答問題，避免冗長和重複內容。\n"
             "User: " + message + "\n"
             "Assistant: "
         )
-        # 呼叫生成管道產生回答
+
+        # Generate response with improved settings
         response = pipe(
             instruction,
-            max_length=200,  # 限制最大輸出長度
+            max_length=1024,  # 限制最大輸出長度，防止無限重複
             truncation=True,
             num_return_sequences=1,
-            temperature=0.3,
-            top_p=0.8,
-            repetition_penalty=1.5,
+            temperature=0.3,  # 保持一定創意但減少胡言亂語
+            top_p=0.8,  # 增加答案多樣性，減少無意義重複
+            repetition_penalty=1.5,  # 降低重複字詞的機率
         )
         generated_text = response[0]['generated_text'].strip()
+
         # 提取 "Assistant:" 之後的部分
         if "Assistant:" in generated_text:
             reply = generated_text.split("Assistant:")[-1].strip()
@@ -45,13 +42,14 @@ def respond(message):
         
         inference_time = time.time() - start_time
         print(f"Inference time: {inference_time:.4f} seconds")
-        return [(message, reply)]
 
-# 定義清空輸入框的函數
-def clear_textbox():
-    return gr.update(value="")
+        return [(message, reply)]
+    
+    except Exception as e:
+        print(f"Error: {e}")
+        return [(message, "Sorry, something went wrong. Please try again.")]
 
-# 設定 Gradio 聊天介面
+# Set up Gradio chat interface
 with gr.Blocks() as demo:
     gr.Markdown("# DeepSeek-R1-Distill-Qwen-1.5B-openvino Chat")
     gr.Markdown("Chat with DeepSeek-R1-Distill-Qwen-1.5B-openvino model.")
@@ -59,8 +57,7 @@ with gr.Blocks() as demo:
     chatbot = gr.Chatbot()
     msg = gr.Textbox(label="Your Message")
     
-    # 當使用者送出訊息時，先觸發 respond() 再清空輸入框
-    msg.submit(respond, inputs=msg, outputs=chatbot).then(clear_textbox, None, msg)
+    msg.submit(respond, msg, chatbot)
 
 if __name__ == "__main__":
     demo.launch(share=True)