Update app.py

app.py

@@ -1,64 +1,18 @@
-import huggingface_hub as hf_hub
-import time
-import openvino_genai as ov_genai
-import numpy as np
 import gradio as gr
-import re
-
-# 下載模型
-model_id = "OpenVINO/Qwen3-0.6B-int4-ov"
-model_path = "Qwen3-0.6B-int4-ov"
-
-hf_hub.snapshot_download(model_id, local_dir=model_path, local_dir_use_symlinks=False)
-
-# 建立推理管線
-device = "CPU"
-pipe = ov_genai.LLMPipeline(model_path, device)
-tokenizer = pipe.get_tokenizer()
-tokenizer.set_chat_template(tokenizer.chat_template)
-
-
-def generate_response(prompt):
-    full_response = ""
-    tokenpersec = ""  # 初始化
-
-    def streamer(subword):
-        nonlocal full_response
-        nonlocal tokenpersec  # 声明 nonlocal
-        full_response += subword
-        yield tokenpersec, full_response  # 使用 yield 逐步更新
-
-    try:
-        pipe.start_chat()
-        generator = pipe.generate(prompt, streamer=streamer, max_new_tokens=1024)  # 建立生成器
-        
-        # 迭代生成器，產生流式更新
-        for tokenpersec, response_chunk in generator:
-            yield tokenpersec, response_chunk # 產生中間更新
-
-        pipe.finish_chat()
-        
-        generated = pipe.generate([prompt], max_length=1024) # 為了得到 perf_metrics
-        tokenpersec=f'{generated.perf_metrics.get_throughput().mean:.2f}'
-
-        yield tokenpersec, full_response  # 產生最終完整更新
+import time
 
-    except Exception as e:
-        yield "發生錯誤", f"生成回應時發生錯誤：{e}"
+def generate_markdown():
+    response = ""
+    for word in ["Hello", "World", "Gradio", "Markdown", "!"]:
+        response += word + " "
+        time.sleep(0.5)
+        yield response
 
+with gr.Blocks() as demo:
+    markdown_output = gr.Markdown(label="回應")
+    button = gr.Button("生成")
 
-# 建立 Gradio 介面
-demo = gr.Interface(
-    fn=generate_response,
-    inputs=gr.Textbox(lines=5, label="輸入提示 (Prompt)"),
-    outputs=[
-        gr.Textbox(label="tokens/sec"),
-        gr.Markdown(label="回應") # 使用 Markdown
-    ],
-    title="Qwen3-0.6B-int4-ov ",
-    description="基於 Qwen3-0.6B-int4-ov 推理應用，支援思考過程分離與 GUI。 使用Markdown顯示"
-)
+    button.click(generate_markdown, outputs=markdown_output)
 
-if __name__ == "__main__":
-    demo.queue()
-    demo.launch()
+demo.queue()
+demo.launch()